diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/system_w0.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/system_wn.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win0.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1636a632453f11c45f638d59f354361052859d98
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win0.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win1.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1e107182dd9c80827155e3ea6452537dd38447fe
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win1.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win2.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c4d3a86ab14a0dd2c5b82bf68fa42b46d9ada59d
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win2.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.470000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.470000–5.000000s: The protagonist turns round and walks back along the route; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win3.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4b8662e66ffa8f9c0c1db45afebed99435847f4e
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win3.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same turn and return stretch; the protagonist moves about half a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,3.833333)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,3.625000)s, [4.416667,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,3.333333)s.
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win4.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..46c27c6869504e357464b77c9817efe8ba93c9fd
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win4.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. Every described entity retains this appearance only while visible in the reference.
+
+The setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same turn and return stretch; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,0.083333)s.
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.666667,2.083333)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win5.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fb7e013113d21588159c195607f9b968d86a8334
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win5.txt
@@ -0,0 +1,31 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.200000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 1.200000–5.000000s: The protagonist walks forward along the route; the protagonist moves about 4 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win6.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0cb7ce75a4978fb15397d4cf51b667dadb9778af
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win6.txt
@@ -0,0 +1,31 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win7.txt b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ceba7c58dd619eb16de2889c7a03b039031c2948
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/colored-obb/win7.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-001-forest-long-return/follow/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5d09c69042ab6bd4d2ae4091b54c930fd151178
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-001-forest-long-return__follow",
+  "scene_number": "001",
+  "camera": "follow",
+  "style": "realistic",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.470000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.470000\u20135.000000s: The protagonist turns round and walks back along the route; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same turn and return stretch; the protagonist moves about half a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same turn and return stretch; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.200000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 1.200000\u20135.000000s: The protagonist walks forward along the route; the protagonist moves about 4 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is winding temperate forest trails, large trunks and fallen timber. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 60-degree vertical field of view. Early-morning low sun streams through the canopy from the left, casting long, soft shadows across the mossy floor and illuminating the scene with a warm, golden glow. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-001-forest-long-return/follow/white/prompts.c2r.json b/prompts/long_video/long30-001-forest-long-return/follow/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..8442f76f3199b736a3302029e2741cee3da8c7a6
--- /dev/null
+++ b/prompts/long_video/long30-001-forest-long-return/follow/white/prompts.c2r.json
@@ -0,0 +1,96 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials. The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "1bb0b575a3159e77d7b51b4b5157453dbc0ddbbeb391baff07313a7bc9d5c279",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 6.25,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials. The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. The protagonist turns round and walks back along the route. Only people visible in the control appear.",
+    "text_sha256": "afc9a271d77fc02cfa26e0211654216ca4bb9af3d395147499bf8e613ec51e3d",
+    "events": [
+      {
+        "t0": 3.75,
+        "t1": 9.97,
+        "window": 1
+      },
+      {
+        "t0": 9.97,
+        "t1": 13.75,
+        "window": 2
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials. The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist turns round and walks back along the route. Only people visible in the control appear.",
+    "text_sha256": "c12925ea5314eee09944d72fd3a2144d4dbb7c67adea7f1cb976537c3bec9941",
+    "events": [
+      {
+        "t0": 9.97,
+        "t1": 13.75,
+        "window": 2
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials. The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist turns round and walks back along the route. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "9db6549a9fb0091114926c1295f3b9276dad6b5abadf74ef8c61804328dbc4c5",
+    "events": [
+      {
+        "t0": 15.0,
+        "t1": 19.95,
+        "window": 4
+      },
+      {
+        "t0": 19.95,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials. The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "1bb0b575a3159e77d7b51b4b5157453dbc0ddbbeb391baff07313a7bc9d5c279",
+    "events": [
+      {
+        "t0": 19.95,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "A realistic temperate woodland in soft daylight filtered through the canopy: rough bark, leaf litter, moss, packed earth trails and weathered timber, with natural human anatomy, believable scale and detailed physical materials. The protagonist is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots. The independent participant B is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "1bb0b575a3159e77d7b51b4b5157453dbc0ddbbeb391baff07313a7bc9d5c279",
+    "events": [
+      {
+        "t0": 19.95,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/system_w0.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/system_wn.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win0.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d07c7e3017f9f5348d0c0bf71d7264d21c532dd8
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win0.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route, rises about half a metre.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route, rises about half a metre.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win1.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..79d5753892c399115186f2fb71146af20a3841f6
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win1.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route, rises about half a metre.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win2.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..964fd497ebc84ebb7195f45bef4dccfb1e1caadc
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win2.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.470000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.470000–5.000000s: The protagonist turns round and walks back along the route; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win3.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7544e3d6dcd1f00b78cd2a05e8b495cc55329031
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win3.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same turn and return stretch; the protagonist moves about half a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,3.500000)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,2.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,3.083333)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win4.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..98d1275357f9d80be0531b3a9f2b7865384dc7f1
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win4.txt
@@ -0,0 +1,31 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+No described person or object is visible in this window. Do not add any.
+
+The setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same turn and return stretch; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win5.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d86c8ce885f9e3dd638b0eb416fb13d2cd2ae067
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win5.txt
@@ -0,0 +1,31 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+No described person or object is visible in this window. Do not add any.
+
+The setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.200000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 1.200000–5.000000s: The protagonist walks forward along the route; the protagonist moves about 4 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win6.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..eafbe79f452e85803d655b5da7599b270cd6b177
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win6.txt
@@ -0,0 +1,31 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+No described person or object is visible in this window. Do not add any.
+
+The setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route, drops a short distance.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win7.txt b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b2442720ef079f054c06ed3a846d4d5877681306
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/colored-obb/win7.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+No described person or object is visible in this window. Do not add any.
+
+The setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-004-mountain-long-return/first-person/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0e638c7efe9efaa94742b4c50b5e4a22d787511
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-004-mountain-long-return__first-person",
+  "scene_number": "004",
+  "camera": "first-person",
+  "style": "ink-wash",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route, rises about half a metre.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route, rises about half a metre.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route, rises about half a metre.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.470000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.470000\u20135.000000s: The protagonist turns round and walks back along the route; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same turn and return stretch; the protagonist moves about half a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nNo described person or object is visible in this window. Do not add any.\n\nThe setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same turn and return stretch; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nNo described person or object is visible in this window. Do not add any.\n\nThe setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.200000s: The protagonist turns round and walks back along the route; the protagonist moves about a metre along its route.; 1.200000\u20135.000000s: The protagonist walks forward along the route; the protagonist moves about 4 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nNo described person or object is visible in this window. Do not add any.\n\nThe setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 3 metres along its route, drops a short distance.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nNo described person or object is visible in this window. Do not add any.\n\nThe setting is a rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Diffuse, shadowless overcast light washes over the scene, creating a flat, even illumination that emphasizes the texture of the ink washes and the separation of the misty depth planes. The whole frame is one finished image in this look: Painterly rendering, long-scroll ink look. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-004-mountain-long-return/first-person/white/prompts.c2r.json b/prompts/long_video/long30-004-mountain-long-return/first-person/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..301753cba97968937cd6debd6a9f4c9ba7296bcf
--- /dev/null
+++ b/prompts/long_video/long30-004-mountain-long-return/first-person/white/prompts.c2r.json
@@ -0,0 +1,96 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "fe51acb7145fb2daa4c7152b24e4fd64b692e33002c7b08a31bb08df327de00f",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 6.25,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. The protagonist turns round and walks back along the route. Only people visible in the control appear.",
+    "text_sha256": "a9e3dc51ee998db1efadd71ba70e9a5b274b6b3fff023f4e3b7134e1e2d4f530",
+    "events": [
+      {
+        "t0": 3.75,
+        "t1": 9.97,
+        "window": 1
+      },
+      {
+        "t0": 9.97,
+        "t1": 13.75,
+        "window": 2
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist turns round and walks back along the route. Only people visible in the control appear.",
+    "text_sha256": "6474bc11b297b8968b6661ee25b7448bf0b9f5b2a0d339ae35744234e1ebf408",
+    "events": [
+      {
+        "t0": 9.97,
+        "t1": 13.75,
+        "window": 2
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist turns round and walks back along the route. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "e79816d90eb3bb48da676f712e767bcc5f2efc986b377209791a408991c9d982",
+    "events": [
+      {
+        "t0": 15.0,
+        "t1": 19.95,
+        "window": 4
+      },
+      {
+        "t0": 19.95,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "fe51acb7145fb2daa4c7152b24e4fd64b692e33002c7b08a31bb08df327de00f",
+    "events": [
+      {
+        "t0": 19.95,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; monochrome black ink with a faint blue-green mineral tint on the peaks; diffuse overcast light; mist separating the ridges into layered depth planes; long sweeping wet brush strokes; most of the background left as empty paper; character design: small figures dwarfed by the surroundings; the whole frame must read as a flat hand-made painting on canvas or paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots. The independent participant C is a hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "fe51acb7145fb2daa4c7152b24e4fd64b692e33002c7b08a31bb08df327de00f",
+    "events": [
+      {
+        "t0": 19.95,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/system_w0.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/system_wn.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win0.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2505512b2d0cdbc59de810e8c2b3af18452c4163
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win0.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win1.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3ef1ea580aafb077112c6ab9c43d177299fa03bf
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win1.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win2.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2991065f7c0755a6a05776794ce6405fb7c6e622
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win2.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win3.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6637ac4ca597068b3ad58ad7a802337d0e38ca4c
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win3.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win4.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..addbcbc93a2b0c4bad60abf2557e102bc4ff9ac9
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win4.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win5.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2991065f7c0755a6a05776794ce6405fb7c6e622
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win5.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win6.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6637ac4ca597068b3ad58ad7a802337d0e38ca4c
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win6.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win7.txt b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..26658afac406c0557eff74efb74779673c4cab84
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/colored-obb/win7.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-005-desert-long-bends/first-person/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..3900e2e7ed469d2f40677a56f3f8e4eb9f55d648
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-005-desert-long-bends__first-person",
+  "scene_number": "005",
+  "camera": "first-person",
+  "style": "realistic",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. \n\nProtagonist action timeline:\n\n0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Preserve the reference spacing, masses and open routes. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Early-morning low sun casts long, soft shadows across the sand, bathing the scene in warm, golden tones that contrast with the cool, hazy blue of the distant sky. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-005-desert-long-bends/first-person/white/prompts.c2r.json b/prompts/long_video/long30-005-desert-long-bends/first-person/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..ffc952a918fdcae9015a774668c4c03c7e0b282f
--- /dev/null
+++ b/prompts/long_video/long30-005-desert-long-bends/first-person/white/prompts.c2r.json
@@ -0,0 +1,86 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "36c6ed025264780d6645c6fcee919d2cfe91853724f7dbc04abe613945e07977",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 6.25,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "36c6ed025264780d6645c6fcee919d2cfe91853724f7dbc04abe613945e07977",
+    "events": [
+      {
+        "t0": 3.75,
+        "t1": 10.0,
+        "window": 1
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "36c6ed025264780d6645c6fcee919d2cfe91853724f7dbc04abe613945e07977",
+    "events": [
+      {
+        "t0": 11.25,
+        "t1": 13.75,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "36c6ed025264780d6645c6fcee919d2cfe91853724f7dbc04abe613945e07977",
+    "events": [
+      {
+        "t0": 15.0,
+        "t1": 21.25,
+        "window": 4
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "36c6ed025264780d6645c6fcee919d2cfe91853724f7dbc04abe613945e07977",
+    "events": [
+      {
+        "t0": 18.75,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The independent participant D is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "36c6ed025264780d6645c6fcee919d2cfe91853724f7dbc04abe613945e07977",
+    "events": [
+      {
+        "t0": 22.5,
+        "t1": 25.0,
+        "window": 6
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/system_w0.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/system_wn.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win0.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1ffe453011a134000e39038ba2eff953fadc08d9
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win0.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.
+
+The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win1.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bb3785b05159320db2f0a439b198455ef99f621b
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win1.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.
+
+The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win2.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..af28875f7919bddbd33929fe0e685077ef93defa
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win2.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.
+
+The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.470000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.470000–5.000000s: The protagonist stops and looks around at the surroundings.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,2.625000)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,4.416667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,2.708333)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win3.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7ee9fe10d68d4576e5595054f22ccd6cce17d275
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win3.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.
+
+The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.700000s: The protagonist stops and looks around at the surroundings.; 1.700000–5.000000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route, lowers the body by a short distance (head height 0.2 to -0.0 m).; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,0.666667)s, [1.708333,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [1.916667,5.166667)s.
+independent participant C (participant-c, identity tag #f736df): visible only during [2.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win4.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bb3785b05159320db2f0a439b198455ef99f621b
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win4.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.
+
+The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win5.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bb3785b05159320db2f0a439b198455ef99f621b
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win5.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.
+
+The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win6.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bb3785b05159320db2f0a439b198455ef99f621b
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win6.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.
+
+The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win7.txt b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7e71ca4499f1b015b4500f8574b368d8cf035333
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/colored-obb/win7.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.
+
+The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-006-snow-long-pause-look/first-person/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..150304dce1840fd4836c71c0f904fc7b2a111403
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-006-snow-long-pause-look__first-person",
+  "scene_number": "006",
+  "camera": "first-person",
+  "style": "fantasy",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed \u2014 frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.\n\nThe independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed \u2014 frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.\n\nThe independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed \u2014 frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.\n\nThe independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.470000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.470000\u20135.000000s: The protagonist stops and looks around at the surroundings.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed \u2014 frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.\n\nThe independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.700000s: The protagonist stops and looks around at the surroundings.; 1.700000\u20135.000000s: The protagonist walks forward along the route; the protagonist moves about 3 metres along its route, lowers the body by a short distance (head height 0.2 to -0.0 m).; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed \u2014 frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.\n\nThe independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed \u2014 frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.\n\nThe independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed \u2014 frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.\n\nThe independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about 2 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about 2 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed \u2014 frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots.\n\nThe independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is snow banks, conifers, faceted ice formations and a frozen open path. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. A low polar sun casts long, soft blue shadows across the snow, illuminating the scene with a cool, desaturated daylight that highlights the texture of the windblown drifts. The whole frame is one finished image in this look: Semi-realistic game rendering. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-006-snow-long-pause-look/first-person/white/prompts.c2r.json b/prompts/long_video/long30-006-snow-long-pause-look/first-person/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..882794f955f8fca3efc71286ff4c5485c9beb8de
--- /dev/null
+++ b/prompts/long_video/long30-006-snow-long-pause-look/first-person/white/prompts.c2r.json
@@ -0,0 +1,96 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots. The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "704384baeacfe11657ec9ac464eea7920b5299cebbdd1a2b301c28de9841c6da",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 6.25,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots. The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. The protagonist stops and looks around at the surroundings. Only people visible in the control appear.",
+    "text_sha256": "81b93af9e084b8f3974545a43a4531a4c14d3d97deb0025d86ba40979dc982be",
+    "events": [
+      {
+        "t0": 3.75,
+        "t1": 9.97,
+        "window": 1
+      },
+      {
+        "t0": 9.97,
+        "t1": 12.95,
+        "window": 2
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots. The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist stops and looks around at the surroundings. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "8b2c2627e90943c718896a8262a755be15d548817e6f10c3bd6088c963f08ca8",
+    "events": [
+      {
+        "t0": 9.97,
+        "t1": 12.95,
+        "window": 2
+      },
+      {
+        "t0": 12.95,
+        "t1": 17.5,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots. The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "704384baeacfe11657ec9ac464eea7920b5299cebbdd1a2b301c28de9841c6da",
+    "events": [
+      {
+        "t0": 12.95,
+        "t1": 21.25,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots. The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "704384baeacfe11657ec9ac464eea7920b5299cebbdd1a2b301c28de9841c6da",
+    "events": [
+      {
+        "t0": 18.75,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "Semi-realistic game rendering: the world, characters and objects keep real-world structure, layout and scale, but proportions, palette, materials and lighting are visibly art-directed — frozen northern saga look: snow, ice and dark pine, fur-and-leather clothing, pale cold light with a faint aurora, rugged slightly idealised figures; an icy steel-blue palette with pale gold highlights; low polar sun raking long blue shadows across the snow; fine windblown snow streaming along the ground; softly painted concept-art surfaces; dramatic sky and light behind the subjects; character design: bulky armoured silhouettes with heavy boots. The independent participant B is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots. The independent participant C is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots. The independent participant D is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "704384baeacfe11657ec9ac464eea7920b5299cebbdd1a2b301c28de9841c6da",
+    "events": [
+      {
+        "t0": 22.5,
+        "t1": 25.0,
+        "window": 6
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/system_w0.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/system_wn.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win0.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d21438a6d6288a5fad39135fc4cfd0e7e1439434
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win0.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win1.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8e87950a8ef33a87e2f7182d9c1d421086cf8fbe
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win1.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win2.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9dbe1723e49ad2e7708a2a8c11acfac68a2fc479
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win2.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.470000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.470000–5.000000s: The protagonist stops and looks around at the surroundings; the protagonist moves a short distance along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,2.583333)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,2.750000)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win3.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..924418aef2539c2b1ad6b5574268e1c5bd3725d1
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win3.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.700000s: The protagonist stops and looks around at the surroundings.; 1.700000–5.000000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [1.875000,5.166667)s.
+independent participant C (participant-c, identity tag #f736df): visible only during [2.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win4.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8e87950a8ef33a87e2f7182d9c1d421086cf8fbe
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win4.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win5.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8e87950a8ef33a87e2f7182d9c1d421086cf8fbe
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win5.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win6.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8e87950a8ef33a87e2f7182d9c1d421086cf8fbe
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win6.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win7.txt b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d4b513b447d48eb3f114356d7078cd85e3af8b12
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/colored-obb/win7.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.
+
+The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+independent participant C (participant-c, identity tag #f736df): visible only during [0.000000,5.166667)s.
+independent participant B (participant-b, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+independent participant D (participant-d, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..9cce6a33ce4ca8fc21194240b24d53f2a89c87c0
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-012-tunnel-long-pause-look__first-person",
+  "scene_number": "012",
+  "camera": "first-person",
+  "style": "realistic",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. \n\nProtagonist action timeline:\n\n0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.470000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.470000\u20135.000000s: The protagonist stops and looks around at the surroundings; the protagonist moves a short distance along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.700000s: The protagonist stops and looks around at the surroundings.; 1.700000\u20135.000000s: The protagonist walks forward along the route; the protagonist moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks forward along the route; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same traverse stretch; the protagonist moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is an underground parking structure with a sloping entrance, pillars, barriers and vehicle lanes. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Shafts of daylight cut through the gloom from high openings, illuminating the dust motes and casting long, sharp shadows across the asphalt floor. An asphalt road runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/white/prompts.c2r.json b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3b72cb95ba9e7b2019346ef59d7a693cb009495
--- /dev/null
+++ b/prompts/long_video/long30-012-tunnel-long-pause-look/first-person/white/prompts.c2r.json
@@ -0,0 +1,96 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "ed585afb1e4d406049c14d6eb32e761a2c4e4dee1f735bda2da8cb5c75771b5d",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 6.25,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. The protagonist stops and looks around at the surroundings. Only people visible in the control appear.",
+    "text_sha256": "fa74ffc28092be250f1cf0873daf7fda7997965d409678f3c4944074f54b9d9e",
+    "events": [
+      {
+        "t0": 3.75,
+        "t1": 9.97,
+        "window": 1
+      },
+      {
+        "t0": 9.97,
+        "t1": 12.95,
+        "window": 2
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist stops and looks around at the surroundings. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "c17363a15eb0b7f2b80aaff59a3484fef2db1149c8185bb36f426eef1b64988e",
+    "events": [
+      {
+        "t0": 9.97,
+        "t1": 12.95,
+        "window": 2
+      },
+      {
+        "t0": 12.95,
+        "t1": 17.5,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "ed585afb1e4d406049c14d6eb32e761a2c4e4dee1f735bda2da8cb5c75771b5d",
+    "events": [
+      {
+        "t0": 12.95,
+        "t1": 21.25,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "ed585afb1e4d406049c14d6eb32e761a2c4e4dee1f735bda2da8cb5c75771b5d",
+    "events": [
+      {
+        "t0": 18.75,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "A realistic built tunnel lit by evenly spaced lamps: poured concrete, steel ribs, cable trays, painted kerbs and a dry floor, with natural human anatomy, believable scale and detailed physical materials. The independent participant B is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independent participant C is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independent participant D is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks forward along the route. Only people visible in the control appear.",
+    "text_sha256": "ed585afb1e4d406049c14d6eb32e761a2c4e4dee1f735bda2da8cb5c75771b5d",
+    "events": [
+      {
+        "t0": 22.5,
+        "t1": 25.0,
+        "window": 6
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/system_w0.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/system_wn.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win0.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..538a48e30d4026fddcbfc163ccbb42c0ee84317a
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win0.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.
+
+The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–3.870000s: The protagonist stands and looks at the nearby landmark; the person pushing the wheeled handcart moves about one and a half metres along its route; the cart moves about one and a half metres along its route.; 3.870000–5.000000s: The protagonist walks away from the landmark; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.000000,4.041667)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,4.041667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.000000,3.958333)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win1.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..65b97c430de8bcb0cfe2a33a9608c0b0a024061c
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win1.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.
+
+The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–0.100000s: The protagonist stands and looks at the nearby landmark.; 0.100000–5.000000s: The protagonist walks away from the landmark; the protagonist moves about 2 metres along its route; the person pushing the wheeled handcart moves about 2 metres along its route; the cart moves about 2 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.000000,0.291667)s, [2.375000,5.166667)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,0.291667)s, [2.583333,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.000000,0.208333)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.000000,4.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win2.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f01591db46e4fe89fee6b3ed7b2f58dc893befdb
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win2.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.
+
+The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks away from the landmark; the protagonist moves about one and a half metres along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same depart stretch; the protagonist moves about a metre along its route; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.000000,2.125000)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,1.916667)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.000000,0.416667)s, [1.583333,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win3.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a2ae122f2b8b67115f64db49cc915e9ca6f9f832
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win3.txt
@@ -0,0 +1,31 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.
+
+The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.600000s: The protagonist walks away from the landmark; the protagonist moves a short distance along its route.; 1.600000–5.000000s: The protagonist pauses at a distance and scans the surroundings.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win4.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8bc989d08e2915b58bdb434ec4e7595f251c2ceb
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win4.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.
+
+The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.770000s: The protagonist pauses at a distance and scans the surroundings; 1.770000–5.000000s: The protagonist walks back toward the landmark; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about one and a half metres along its route; the cart moves about one and a half metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.000000,2.000000)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [2.250000,5.166667)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [2.333333,5.166667)s.
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [2.375000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win5.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c6e09ac17cc517cbd2d695ab39d21b7a53f0e69d
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win5.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.
+
+The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks back toward the landmark; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same return stretch; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.000000,2.333333)s, [3.916667,5.166667)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,2.500000)s, [3.666667,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win6.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b54f8740df909eeab34bd5c92c0b8f8d51c95e0d
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win6.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.
+
+The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks back toward the landmark; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same return stretch; the protagonist moves a short distance along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.166667,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win7.txt b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0e988d04f2c25573ca73fc35359576c16e95b9ee
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/colored-obb/win7.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.
+
+The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-016-urban-park-revisit/first-person/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..98a0acb1f0b12150b115ffae368a15b6fc7e645f
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-016-urban-park-revisit__first-person",
+  "scene_number": "016",
+  "camera": "first-person",
+  "style": "realistic",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.\n\nThe wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20133.870000s: The protagonist stands and looks at the nearby landmark; the person pushing the wheeled handcart moves about one and a half metres along its route; the cart moves about one and a half metres along its route.; 3.870000\u20135.000000s: The protagonist walks away from the landmark; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.\n\nThe wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20130.100000s: The protagonist stands and looks at the nearby landmark.; 0.100000\u20135.000000s: The protagonist walks away from the landmark; the protagonist moves about 2 metres along its route; the person pushing the wheeled handcart moves about 2 metres along its route; the cart moves about 2 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.\n\nThe wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks away from the landmark; the protagonist moves about one and a half metres along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same depart stretch; the protagonist moves about a metre along its route; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.\n\nThe independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.600000s: The protagonist walks away from the landmark; the protagonist moves a short distance along its route.; 1.600000\u20135.000000s: The protagonist pauses at a distance and scans the surroundings.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.\n\nThe wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.770000s: The protagonist pauses at a distance and scans the surroundings; 1.770000\u20135.000000s: The protagonist walks back toward the landmark; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about one and a half metres along its route; the cart moves about one and a half metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.\n\nThe wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks back toward the landmark; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same return stretch; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.\n\nThe wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks back toward the landmark; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same return stretch; the protagonist moves a short distance along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials.\n\nThe wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a landscaped public park with branching paths, benches, trees and a paved plaza. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. The overcast sky provides diffuse, shadowless illumination with a cool colour temperature, creating a calm and muted atmosphere. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-016-urban-park-revisit/first-person/white/prompts.c2r.json b/prompts/long_video/long30-016-urban-park-revisit/first-person/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..c57742431b9fc9605f43c241b620a71ab9b191a4
--- /dev/null
+++ b/prompts/long_video/long30-016-urban-park-revisit/first-person/white/prompts.c2r.json
@@ -0,0 +1,101 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist stands and looks at the nearby landmark. The protagonist walks away from the landmark. Only people visible in the control appear.",
+    "text_sha256": "9a092062d8ad2b96cdc3b58df8d75a52ff74780ea796c9c356761e64e4aec94b",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 3.87,
+        "window": 0
+      },
+      {
+        "t0": 3.85,
+        "t1": 10.0,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks away from the landmark. Only people visible in the control appear.",
+    "text_sha256": "9c236d4a734126c0c95547e183ee71bd31de70de7d4a36528d0e2619eea2b917",
+    "events": [
+      {
+        "t0": 3.85,
+        "t1": 10.0,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks away from the landmark. The protagonist pauses at a distance and scans the surroundings. Only people visible in the control appear.",
+    "text_sha256": "15c120a17b4bf0fb9e487df2ab897886dc6a7e71314a3f0f008715d5d50bedeb",
+    "events": [
+      {
+        "t0": 11.25,
+        "t1": 12.85,
+        "window": 3
+      },
+      {
+        "t0": 12.85,
+        "t1": 16.77,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist pauses at a distance and scans the surroundings. The protagonist walks back toward the landmark. Only people visible in the control appear.",
+    "text_sha256": "2e31a834bacd38d4fc741e273e84881698ae54e08f635ec0efce5f4edb8dfbf1",
+    "events": [
+      {
+        "t0": 12.85,
+        "t1": 16.77,
+        "window": 3
+      },
+      {
+        "t0": 16.77,
+        "t1": 21.25,
+        "window": 4
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks back toward the landmark. Only people visible in the control appear.",
+    "text_sha256": "e6dde3c2b90bfaa041712d8b60244dcac6fb20011a0e49bd2be210a6f17a7635",
+    "events": [
+      {
+        "t0": 16.77,
+        "t1": 25.0,
+        "window": 4
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "A realistic city park in soft daylight: mown lawn, gravel paths, benches, clipped hedges, mature trees and low railings, with natural human anatomy, believable scale and detailed physical materials. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels. The independently moving adult participant 1 is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The independently moving adult participant 2 is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals. The person pushing the wheeled handcart is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks back toward the landmark. Only people visible in the control appear.",
+    "text_sha256": "e6dde3c2b90bfaa041712d8b60244dcac6fb20011a0e49bd2be210a6f17a7635",
+    "events": [
+      {
+        "t0": 22.5,
+        "t1": 25.0,
+        "window": 6
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/system_w0.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/system_wn.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/win0.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c7535e59dfa5ddceede58a0ed290a4b9251c2e30
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win0.txt
@@ -0,0 +1,35 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–3.870000s: The protagonist stands on the route and takes in the surroundings; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 3.870000–5.000000s: The protagonist walks along the route while looking around at the surroundings.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/win1.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0e9d4a1201dd797adcf1127fc4753bf9b32eab05
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win1.txt
@@ -0,0 +1,35 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–0.100000s: The protagonist stands on the route and takes in the surroundings.; 0.100000–5.000000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,1.916667)s.
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.000000,2.250000)s, [2.583333,3.291667)s.
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.000000,1.958333)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.000000,1.916667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/win2.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fc99e96fcb2fe0aa2aafc8b61ac5e39ae13a9bcb
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win2.txt
@@ -0,0 +1,31 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same camera exploration stretch; the protagonist moves about a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/win3.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..06530f93feff04a09c0c0ce5fcb5c1a12b7c6267
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win3.txt
@@ -0,0 +1,35 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route; 2.500000–5.000000s: The protagonist keeps going through the same camera exploration stretch; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [3.833333,5.166667)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [4.291667,5.166667)s.
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [4.333333,5.166667)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [4.375000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/win4.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6dcced7e39297e4db4e4ea65256afbb068bafb94
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win4.txt
@@ -0,0 +1,35 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same camera exploration stretch; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.083333,2.500000)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.541667,3.375000)s.
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.583333,2.791667)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.625000,2.750000)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/win5.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6a42ae1382e6d2a1871a7e017844bda45e3162be
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win5.txt
@@ -0,0 +1,31 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route.; 2.500000–5.000000s: The protagonist keeps going through the same camera exploration stretch; the protagonist moves about half a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/win6.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9c81067eb87ee39fccb1153ebb66c92b4dd159d0
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win6.txt
@@ -0,0 +1,35 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.470000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about half a metre along its route; 1.470000–5.000000s: The protagonist turns back to face forward along the route; the protagonist moves about half a metre along its route; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [2.166667,5.166667)s.
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [3.041667,5.166667)s.
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [3.083333,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [3.583333,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-017-home-camera/follow/colored-obb/win7.txt b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6f307f58ae2e38b02d970b10a8d93c9a44735369
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/colored-obb/win7.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.
+
+The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+a wheeled handcart (ambient-cart, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #366ef7): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-01, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-03, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an independently moving adult participant (npc-02, identity tag #f736df): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-017-home-camera/follow/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-017-home-camera/follow/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..96b2e6bda680f8c3e0e6364db3b1e1bd9d85d53b
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-017-home-camera__follow",
+  "scene_number": "017",
+  "camera": "follow",
+  "style": "sketch",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20133.870000s: The protagonist stands on the route and takes in the surroundings; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 3.870000\u20135.000000s: The protagonist walks along the route while looking around at the surroundings.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20130.100000s: The protagonist stands on the route and takes in the surroundings.; 0.100000\u20135.000000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about a metre along its route; the cart moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same camera exploration stretch; the protagonist moves about a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route; 2.500000\u20135.000000s: The protagonist keeps going through the same camera exploration stretch; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same camera exploration stretch; the protagonist moves about a metre along its route; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about a metre along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same camera exploration stretch; the protagonist moves about half a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.470000s: The protagonist walks along the route while looking around at the surroundings; the protagonist moves about half a metre along its route; 1.470000\u20135.000000s: The protagonist turns back to face forward along the route; the protagonist moves about half a metre along its route; the person pushing the wheeled handcart moves about half a metre along its route; the cart moves about half a metre along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters.\n\nThe protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a fully roofed furnished home with connected rooms, cabinets, chairs and tables. Preserve the reference spacing, masses and open routes. A steel utility cart with a flat bed and rubber wheels. The camera follows the recorded third-person path of the conditioning reference with its fixed 53-degree vertical field of view. Warm tungsten lamps mixed with window daylight cast soft, directional light from the left, creating gentle shadows and highlighting dust motes floating in the air. The whole frame is one finished image in this look: Graphic non-photorealistic rendering, coloured pencil sketch look. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-017-home-camera/follow/white/prompts.c2r.json b/prompts/long_video/long30-017-home-camera/follow/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..e223f70403bc81f60f62f4001044a7b7c83a8152
--- /dev/null
+++ b/prompts/long_video/long30-017-home-camera/follow/white/prompts.c2r.json
@@ -0,0 +1,96 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist stands on the route and takes in the surroundings. The protagonist walks along the route while looking around at the surroundings. Only people visible in the control appear.",
+    "text_sha256": "619feb4694650bf2462d0ae25944418586e582bcab879eaa9c0b5cc19aae1f45",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 3.87,
+        "window": 0
+      },
+      {
+        "t0": 3.85,
+        "t1": 10.0,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks along the route while looking around at the surroundings. Only people visible in the control appear.",
+    "text_sha256": "e78bf1fc3a522b6c2dd0d1d2e954095df667d91b7af46781f1bf863e8e99d5d2",
+    "events": [
+      {
+        "t0": 3.85,
+        "t1": 10.0,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks along the route while looking around at the surroundings. Only people visible in the control appear.",
+    "text_sha256": "e78bf1fc3a522b6c2dd0d1d2e954095df667d91b7af46781f1bf863e8e99d5d2",
+    "events": [
+      {
+        "t0": 11.25,
+        "t1": 13.75,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks along the route while looking around at the surroundings. Only people visible in the control appear.",
+    "text_sha256": "e78bf1fc3a522b6c2dd0d1d2e954095df667d91b7af46781f1bf863e8e99d5d2",
+    "events": [
+      {
+        "t0": 15.0,
+        "t1": 21.25,
+        "window": 4
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist walks along the route while looking around at the surroundings. The protagonist turns back to face forward along the route. Only people visible in the control appear.",
+    "text_sha256": "97b2b7d02de1a356232867d7a7ab9ba2efc27ae47873a25d78e5d67d4c7ef470",
+    "events": [
+      {
+        "t0": 18.75,
+        "t1": 23.97,
+        "window": 5
+      },
+      {
+        "t0": 23.97,
+        "t1": 27.5,
+        "window": 6
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; soft coloured-pencil hues of teal, peach and warm brown on cream paper; cosy window light from one side of the room; visible paper tooth under the pencil strokes; layered scribbled tone; background fading out to blank paper at the edges; character design: figures in quick gesture lines; the whole frame must read as a flat hand-made drawing or print on paper with visible medium texture - not a smooth 3D render, not a game screenshot, no cel-shaded 3D characters. The protagonist is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers. The wheeled handcart is a steel utility cart with a flat bed and rubber wheels, sketched in slate grey with subtle shading to indicate its metallic surface. The independently moving adult participant 1 is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes. The independently moving adult participant 2 is a chef with dark hair under a cap, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs. The person pushing the wheeled handcart is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots. Every described entity retains this appearance only while visible in the reference. The protagonist turns back to face forward along the route. Only people visible in the control appear.",
+    "text_sha256": "d99d6d09e8326c83f1c739eeda6387ab5234e4220fa28346aab763f9861924a3",
+    "events": [
+      {
+        "t0": 23.97,
+        "t1": 27.5,
+        "window": 6
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/system_w0.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/system_wn.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win0.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1726ca1e1cde64141bf4957197c7503a9cdd4e63
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win0.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.
+
+The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–3.570000s: The protagonist raises the training firearm and aims at the training opponents; the protagonist moves about half a metre along its route; 3.570000–5.000000s: The protagonist fires at the training opponents while they return fire.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a second training opponent returning fire from the opposite side (opponent-b, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+a training opponent holding position and returning fire (opponent-a, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win1.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..cfd4082e29902282f91933598901c25c4fa2d310
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win1.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.
+
+The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist fires at the training opponents while they return fire.; 2.500000–5.000000s: The protagonist keeps going through the same exchange stretch; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a second training opponent returning fire from the opposite side (opponent-b, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+a training opponent holding position and returning fire (opponent-a, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win2.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4127a6dc98dd43ed12d1382a4694588a5be5f722
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win2.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.
+
+The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.070000s: The protagonist fires at the training opponents while they return fire.; 2.070000–5.000000s: The protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.2 m), bends the knees deeply; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a second training opponent returning fire from the opposite side (opponent-b, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+a training opponent holding position and returning fire (opponent-a, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win3.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4a6171370e90a52881038d57efccca09d82fe743
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win3.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.
+
+The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.900000s: The protagonist runs to the nearby cover; the protagonist moves about a metre along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; 1.900000–5.000000s: The protagonist crouches behind the cover and reloads the training firearm.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a second training opponent returning fire from the opposite side (opponent-b, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+a training opponent holding position and returning fire (opponent-a, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win4.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f59ad8ec99fbef8f874fbe2ea27c7052dc447f5b
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win4.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.
+
+The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.370000s: The protagonist crouches behind the cover and reloads the training firearm.; 2.370000–5.000000s: The protagonist rises from behind the cover and steps back out; the protagonist moves about one and a half metres along its route, rises to full height (head height 0.9 to 1.6 m), lowers the hands, straightens the legs; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a second training opponent returning fire from the opposite side (opponent-b, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+a training opponent holding position and returning fire (opponent-a, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win5.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6887e612d264238aeeda1e19483d33112c274f22
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win5.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.
+
+The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.900000s: The protagonist rises from behind the cover and steps back out; the protagonist moves about a metre along its route, straightens the legs; 1.900000–5.000000s: The protagonist fires at the training opponents again while they return fire; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a second training opponent returning fire from the opposite side (opponent-b, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+a training opponent holding position and returning fire (opponent-a, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win6.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9e659c26e9f8125ad8ec1d0fb18bd2a1c4838128
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win6.txt
@@ -0,0 +1,32 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.
+
+The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist fires at the training opponents again while they return fire.; 2.500000–5.000000s: The protagonist keeps going through the same second exchange stretch.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+a second training opponent returning fire from the opposite side (opponent-b, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+a training opponent holding position and returning fire (opponent-a, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win7.txt b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d640879a5659e7328f26aabdd5a38485cddcaf03
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/colored-obb/win7.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.
+
+The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+a second training opponent returning fire from the opposite side (opponent-b, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+a training opponent holding position and returning fire (opponent-a, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-019-warehouse-combat/first-person/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8a865411541b7acd8d5a6f5cc29b456fccaaf98
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-019-warehouse-combat__first-person",
+  "scene_number": "019",
+  "camera": "first-person",
+  "style": "anime",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world \u2014 dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.\n\nThe training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20133.570000s: The protagonist raises the training firearm and aims at the training opponents; the protagonist moves about half a metre along its route; 3.570000\u20135.000000s: The protagonist fires at the training opponents while they return fire.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world \u2014 dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.\n\nThe training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist fires at the training opponents while they return fire.; 2.500000\u20135.000000s: The protagonist keeps going through the same exchange stretch; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world \u2014 dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.\n\nThe training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.070000s: The protagonist fires at the training opponents while they return fire.; 2.070000\u20135.000000s: The protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.2 m), bends the knees deeply; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world \u2014 dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.\n\nThe training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.900000s: The protagonist runs to the nearby cover; the protagonist moves about a metre along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; 1.900000\u20135.000000s: The protagonist crouches behind the cover and reloads the training firearm.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world \u2014 dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.\n\nThe training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.370000s: The protagonist crouches behind the cover and reloads the training firearm.; 2.370000\u20135.000000s: The protagonist rises from behind the cover and steps back out; the protagonist moves about one and a half metres along its route, rises to full height (head height 0.9 to 1.6 m), lowers the hands, straightens the legs; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world \u2014 dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.\n\nThe training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.900000s: The protagonist rises from behind the cover and steps back out; the protagonist moves about a metre along its route, straightens the legs; 1.900000\u20135.000000s: The protagonist fires at the training opponents again while they return fire; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world \u2014 dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.\n\nThe training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist fires at the training opponents again while they return fire.; 2.500000\u20135.000000s: The protagonist keeps going through the same second exchange stretch.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world \u2014 dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot.\n\nThe training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a roofed warehouse with tall stocked racks, wide cross aisles and crates. Preserve the reference spacing, masses and open routes. The camera is the protagonist's own first-person view, carried at the protagonist's eye height, and it follows the recorded camera path of the conditioning reference with its fixed 75-degree vertical field of view. Cold overhead strip lights cast harsh highlights and deep pools of shadow, with visible dust motes hanging in the light shafts. The whole frame is one finished image in this look: Anime-style rendering. There are no trees in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-019-warehouse-combat/first-person/white/prompts.c2r.json b/prompts/long_video/long30-019-warehouse-combat/first-person/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b77500079438465224b3b327c05a0596b1ee2d1
--- /dev/null
+++ b/prompts/long_video/long30-019-warehouse-combat/first-person/white/prompts.c2r.json
@@ -0,0 +1,111 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot. The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference. The protagonist raises the training firearm and aims at the training opponents. The protagonist fires at the training opponents while they return fire. Only people visible in the control appear.",
+    "text_sha256": "81cea51ff863c185bf6b0ab51b71bcc40e706c873dee25a31db7349dd283529d",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 3.57,
+        "window": 0
+      },
+      {
+        "t0": 3.57,
+        "t1": 6.25,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot. The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference. The protagonist fires at the training opponents while they return fire. The protagonist runs to the nearby cover. Only people visible in the control appear.",
+    "text_sha256": "1e7805c1b2070529e6698547376fa0f7e13896487666d3c3b35be302b7b0feea",
+    "events": [
+      {
+        "t0": 3.57,
+        "t1": 9.57,
+        "window": 0
+      },
+      {
+        "t0": 9.57,
+        "t1": 13.15,
+        "window": 2
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot. The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference. The protagonist runs to the nearby cover. The protagonist crouches behind the cover and reloads the training firearm. Only people visible in the control appear.",
+    "text_sha256": "488e078c2c5ac67eb5eb29e859c2584858f3c7dc2f7273752f02686cf238a974",
+    "events": [
+      {
+        "t0": 9.57,
+        "t1": 13.15,
+        "window": 2
+      },
+      {
+        "t0": 13.15,
+        "t1": 17.37,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot. The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference. The protagonist crouches behind the cover and reloads the training firearm. The protagonist rises from behind the cover and steps back out. Only people visible in the control appear.",
+    "text_sha256": "336aa4877271903aa1feb71e19cef0233e9040aa414644f74a6af6f04e83e075",
+    "events": [
+      {
+        "t0": 13.15,
+        "t1": 17.37,
+        "window": 3
+      },
+      {
+        "t0": 17.37,
+        "t1": 20.65,
+        "window": 4
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot. The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference. The protagonist rises from behind the cover and steps back out. The protagonist fires at the training opponents again while they return fire. Only people visible in the control appear.",
+    "text_sha256": "db4a3cebbd7d6e78e5906d73d5ec522ba3700a25c3eaab3600d00e4f69e76cee",
+    "events": [
+      {
+        "t0": 17.37,
+        "t1": 20.65,
+        "window": 4
+      },
+      {
+        "t0": 20.65,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — dark gritty anime look: heavy black ink shadows, thick contour lines, muted desaturated colours with harsh highlights, grainy textured backgrounds; a desaturated olive-and-rust palette; cold overhead strip lights with deep pools of shadow; dust hanging in the light shafts; slightly wobbly hand-inked outlines; detailed line-drawn backgrounds with light washes; character design: compact stocky figures with thick outlines; strictly a flat hand-drawn 2D anime frame with line art - not a 3D render, not a game screenshot. The training rifle with stock, receiver, barrel and magazine is a matte black training rifle with a stock, receiver, barrel, and magazine, held by the protagonist. The training opponent holding position and returning fire is a stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a matte black training firearm. The second training opponent returning fire from the opposite side is a stocky adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, holding a matte black training firearm. Every described entity retains this appearance only while visible in the reference. The protagonist fires at the training opponents again while they return fire. Only people visible in the control appear.",
+    "text_sha256": "559bb37eec182b6e4c6ba9da9145e3d7072f98adb51ce7e1850be544bfb61e6c",
+    "events": [
+      {
+        "t0": 20.65,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/system_w0.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/system_wn.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win0.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1a376bac691000cd55671ebce93ed3362aed5782
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win0.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–0.900000s: The protagonist's sedan accelerates away from the start while the rival sedans keep racing on their own lines; the protagonist moves about 3 metres along its route; the vehicle moves about 3 metres along its route; the vehicle moves about 2 metres along its route; the vehicle moves about a metre along its route.; 0.900000–5.000000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 62 metres along its route; the vehicle moves about 62 metres along its route; the vehicle moves about 66 metres along its route; the vehicle moves about 71 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win1.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e93746ca6f3911a1d84286660dad10fd668fa21c
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win1.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 32 metres along its route; the vehicle moves about 32 metres along its route; the vehicle moves about 37 metres along its route; the vehicle moves about 40 metres along its route.; 1.500000–2.770000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 35 metres along its route; the vehicle moves about 35 metres along its route; the vehicle moves about 40 metres along its route; the vehicle moves about 40 metres along its route.; 2.770000–5.000000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 70 metres along its route; the vehicle moves about 70 metres along its route; the vehicle moves about 83 metres along its route; the vehicle moves about 73 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win2.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..edc22d80bbc244e11f7ce7cc98344dec1280d439
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win2.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 83 metres along its route; the vehicle moves about 83 metres along its route; the vehicle moves about 94 metres along its route; the vehicle moves about 82 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 88 metres along its route; the vehicle moves about 88 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 82 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win3.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..cb5542de917fd15fef479c4b86a22fb138eadf2c
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win3.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 89 metres along its route; the vehicle moves about 89 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 83 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 90 metres along its route; the vehicle moves about 90 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 85 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win4.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2cb66db73c49c289ee1776218efa5d3d5bff2920
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win4.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 91 metres along its route; the vehicle moves about 91 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 86 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 87 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win5.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..21e0194799a25e4b3a2918d8755b7399d44620a1
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win5.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 87 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 88 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win6.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..95d65cffbcb367949bd4a31266833eb34a9ee8af
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win6.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 88 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 89 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win7.txt b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..66bae3621d3d740e3b1cd61eb140b3cf3b838fee
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/colored-obb/win7.txt
@@ -0,0 +1,37 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-021-desert-launch/follow/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-021-desert-launch/follow/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..af476f18a5af6a9da786f8da08f6ee36616dcc0f
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-021-desert-launch__follow",
+  "scene_number": "021",
+  "camera": "follow",
+  "style": "realistic",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20130.900000s: The protagonist's sedan accelerates away from the start while the rival sedans keep racing on their own lines; the protagonist moves about 3 metres along its route; the vehicle moves about 3 metres along its route; the vehicle moves about 2 metres along its route; the vehicle moves about a metre along its route.; 0.900000\u20135.000000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 62 metres along its route; the vehicle moves about 62 metres along its route; the vehicle moves about 66 metres along its route; the vehicle moves about 71 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 32 metres along its route; the vehicle moves about 32 metres along its route; the vehicle moves about 37 metres along its route; the vehicle moves about 40 metres along its route.; 1.500000\u20132.770000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 35 metres along its route; the vehicle moves about 35 metres along its route; the vehicle moves about 40 metres along its route; the vehicle moves about 40 metres along its route.; 2.770000\u20135.000000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 70 metres along its route; the vehicle moves about 70 metres along its route; the vehicle moves about 83 metres along its route; the vehicle moves about 73 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 83 metres along its route; the vehicle moves about 83 metres along its route; the vehicle moves about 94 metres along its route; the vehicle moves about 82 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 88 metres along its route; the vehicle moves about 88 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 82 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 89 metres along its route; the vehicle moves about 89 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 83 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 90 metres along its route; the vehicle moves about 90 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 85 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 91 metres along its route; the vehicle moves about 91 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 86 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 87 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 87 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 88 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 88 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same follow stretch; the protagonist moves about 92 metres along its route; the vehicle moves about 92 metres along its route; the vehicle moves about 93 metres along its route; the vehicle moves about 89 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is sand dunes, angular desert rocks and cactus plants. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The camera follows the recorded third-person path of the conditioning reference. The dawn-low-sun lighting creates a warm, golden hue with long, soft shadows stretching across the sand, emphasizing the textures of the rocks and the ripples in the sand. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. There are no trees in this clip. There are no buildings or ruins in this clip. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-021-desert-launch/follow/white/prompts.c2r.json b/prompts/long_video/long30-021-desert-launch/follow/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca7e06502af9803686899da08b257a230e3fd8f9
--- /dev/null
+++ b/prompts/long_video/long30-021-desert-launch/follow/white/prompts.c2r.json
@@ -0,0 +1,91 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan accelerates away from the start while the rival sedans keep racing on their own lines. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "578bd3ddf88b1806387febbb85b57ab78c10a7cde092470077dfcbebe8c236ce",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 0.9,
+        "window": 0
+      },
+      {
+        "t0": 0.9,
+        "t1": 10.0,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "2d1213ec395c3106e3c65be9f6e3f34014047d1d3c8a1bbfc9924b52c499b6b3",
+    "events": [
+      {
+        "t0": 0.9,
+        "t1": 10.0,
+        "window": 0
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "2d1213ec395c3106e3c65be9f6e3f34014047d1d3c8a1bbfc9924b52c499b6b3",
+    "events": [
+      {
+        "t0": 11.25,
+        "t1": 13.75,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "2d1213ec395c3106e3c65be9f6e3f34014047d1d3c8a1bbfc9924b52c499b6b3",
+    "events": [
+      {
+        "t0": 15.0,
+        "t1": 21.25,
+        "window": 4
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "2d1213ec395c3106e3c65be9f6e3f34014047d1d3c8a1bbfc9924b52c499b6b3",
+    "events": [
+      {
+        "t0": 18.75,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "A realistic sun-baked desert in strong daylight: rippled sand, pale rock, dry scrub and heat-softened horizons, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels, with a seated driver. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the nearer rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "2d1213ec395c3106e3c65be9f6e3f34014047d1d3c8a1bbfc9924b52c499b6b3",
+    "events": [
+      {
+        "t0": 22.5,
+        "t1": 25.0,
+        "window": 6
+      }
+    ]
+  }
+]
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/system_w0.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/system_w0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d114fed95eb00f2a8001b5ccedc29d8fbbc31e1c
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/system_w0.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. <Picture 1> supplies appearance and the locked initial frame. Start at that state and follow the current reference throughout this window.
\ No newline at end of file
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/system_wn.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/system_wn.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8aff76988a29669a5ccef09317cb638ed8f39c0
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/system_wn.txt
@@ -0,0 +1 @@
+AWM_PROXY_CONTROL. <Video 1> controls exact current camera, geometry, subject visibility, root motion and facing. Proxy colors are identity tags, not target surfaces. Materialize each visible entity once in its occupied region. No proxy boxes, outlines, white front marks or interface overlays. No appearance image is supplied in this continuation. Preserve identity from the generated prefix. The first34 frames are the locked end of the preceding window; continue without restarting, freezing or returning to an old camera composition.
\ No newline at end of file
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win0.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1f3a3e5c59a91c5bf53c000a758aa9ed0aa203b7
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win0.txt
@@ -0,0 +1,38 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. This is the opening window.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 is a diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels, the same wetsuit, and fins, seated in the driver's seat of the rival sedan whenever that driver is visible in the reference. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–2.630000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 59 metres along its route; the vehicle moves about 59 metres along its route; the vehicle moves about 60 metres along its route; the vehicle moves about 85 metres along its route; the vehicle moves about 62 metres along its route.; 2.630000–5.000000s: The protagonist's sedan pulls out alongside a rival sedan and tries to pass; the protagonist moves about 63 metres along its route; the vehicle moves about 63 metres along its route; the vehicle moves about 55 metres along its route; the vehicle moves about 77 metres along its route; the vehicle moves about 49 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,1.541667)s, [1.583333,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-2, identity tag #f79e36): visible only during [0.000000,1.541667)s, [1.583333,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,1.458333)s, [1.583333,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,1.541667)s, [1.583333,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,1.541667)s, [1.583333,5.166667)s.
+the seated driver of rival racing car 2 (driver-2, identity tag #f73677): visible only during [0.000000,1.541667)s, [1.583333,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-3, identity tag #6636f7): visible only during [0.625000,3.125000)s.
+the seated driver of rival racing car 3 (driver-3, identity tag #36aff7): visible only during [1.083333,1.541667)s, [1.583333,2.875000)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win1.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win1.txt
new file mode 100644
index 0000000000000000000000000000000000000000..daffee842e1e674959e57f400f235ad36a4fca3b
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win1.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–0.630000s: The protagonist's sedan pulls out alongside a rival sedan and tries to pass; the protagonist moves about 17 metres along its route; the vehicle moves about 17 metres along its route; the vehicle moves about 14 metres along its route; the vehicle moves about 20 metres along its route.; 0.630000–2.670000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 60 metres along its route; the vehicle moves about 60 metres along its route; the vehicle moves about 48 metres along its route; the vehicle moves about 64 metres along its route.; 2.670000–3.930000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 38 metres along its route; the vehicle moves about 38 metres along its route; the vehicle moves about 30 metres along its route; the vehicle moves about 42 metres along its route.; 3.930000–5.000000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 36 metres along its route; the vehicle moves about 36 metres along its route; the vehicle moves about 26 metres along its route; the vehicle moves about 36 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,5.000000)s.
+an AI-driven rival sedan with a seated driver (rival-2, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 2 (driver-2, identity tag #f73677): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win2.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win2.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4381671bfdd71aa05d153cb1c0f54bfd6e4b3df5
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win2.txt
@@ -0,0 +1,36 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–0.200000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 5 metres along its route; the vehicle moves about 5 metres along its route; the vehicle moves about 4 metres along its route; the vehicle moves about 6 metres along its route.; 0.200000–3.570000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 117 metres along its route; the vehicle moves about 118 metres along its route; the vehicle moves about 84 metres along its route; the vehicle moves about 113 metres along its route.; 3.570000–5.000000s: The protagonist's sedan brakes for the corner and turns in; the protagonist moves about 47 metres along its route; the vehicle moves about 47 metres along its route; the vehicle moves about 35 metres along its route; the vehicle moves about 47 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+the seated driver of rival racing car 1 (driver-1, identity tag #3ef736): visible only during [0.000000,1.250000)s.
+an AI-driven rival sedan with a seated driver (rival-2, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+an AI-driven rival sedan with a seated driver (rival-1, identity tag #36f7d6): visible only during [0.000000,1.541667)s.
+the seated driver of rival racing car 2 (driver-2, identity tag #f73677): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win3.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win3.txt
new file mode 100644
index 0000000000000000000000000000000000000000..35436828a0faa7106c5f2cf081b5f0b70b24845a
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win3.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–1.070000s: The protagonist's sedan brakes for the corner and turns in; the protagonist moves about 35 metres along its route; the vehicle moves about 35 metres along its route; the vehicle moves about 35 metres along its route.; 1.070000–3.530000s: The protagonist's sedan pulls out alongside a rival sedan and tries to pass; the protagonist moves about 95 metres along its route; the vehicle moves about 95 metres along its route; the vehicle moves about 84 metres along its route.; 3.530000–5.000000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 61 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 50 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an AI-driven rival sedan with a seated driver (rival-2, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 2 (driver-2, identity tag #f73677): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win4.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win4.txt
new file mode 100644
index 0000000000000000000000000000000000000000..74decc80cee505e68158b561165e253b887d8563
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win4.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 105 metres along its route; the vehicle moves about 105 metres along its route; the vehicle moves about 86 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same overtaking stretch; the protagonist moves about 102 metres along its route; the vehicle moves about 102 metres along its route; the vehicle moves about 86 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an AI-driven rival sedan with a seated driver (rival-2, identity tag #f79e36): visible only during [0.000000,5.166667)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 2 (driver-2, identity tag #f73677): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win5.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win5.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6f73b8651c5522b71243d2a0cd61da7db319d30c
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win5.txt
@@ -0,0 +1,34 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–3.400000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 136 metres along its route; the vehicle moves about 136 metres along its route; the vehicle moves about 117 metres along its route.; 3.400000–5.000000s: The protagonist's sedan cruises along the track at steady speed while the rival sedans keep racing on their own lines; the protagonist moves about 62 metres along its route; the vehicle moves about 62 metres along its route; the vehicle moves about 55 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an AI-driven rival sedan with a seated driver (rival-2, identity tag #f79e36): visible only during [0.000000,4.125000)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+the seated driver of rival racing car 2 (driver-2, identity tag #f73677): visible only during [0.000000,3.666667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win6.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win6.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f9ca1cdc55730620c28c9f7705bb6f6babf6438f
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win6.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000–2.500000s: The protagonist's sedan cruises along the track at steady speed while the rival sedans keep racing on their own lines; the protagonist moves about 98 metres along its route; the vehicle moves about 98 metres along its route; the vehicle moves about 88 metres along its route.; 2.500000–5.000000s: The protagonist keeps going through the same cruise stretch; the protagonist moves about 98 metres along its route; the vehicle moves about 98 metres along its route; the vehicle moves about 90 metres along its route.; 5.000000–5.166667s: Everyone and the camera simply keep moving as before.
+
+Exact current reference visibility (local half-open intervals):
+an AI-driven rival sedan with a seated driver (rival-2, identity tag #f79e36): visible only during [0.000000,0.375000)s.
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win7.txt b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win7.txt
new file mode 100644
index 0000000000000000000000000000000000000000..95c5c0fc0cdd3080e3c274edf96512a4ee484c49
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/colored-obb/win7.txt
@@ -0,0 +1,33 @@
+subject_definitions:
+<Video 1> is the structural spatial and temporal reference for the current window.
+
+summary:
+[reference generation] Render a continuous 5.166667-second window in the stated target style. Continue the ongoing take from its already given first 34 frames; do not restart.
+
+retention_analysis:
+<Video 1>: partially_preserved — retain geometry, camera, visible count, position, scale, facing and action timing; render the target appearance and materials.
+
+detailed_description:
+[Shot 1] Appearance and scene:
+
+Rendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.
+
+The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. Every described entity retains this appearance only while visible in the reference.
+
+The setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. There is no box, frame, glass panel, outline or coloured slab around or in front of any person, animal or object in the output; every figure stands directly in the scene.
+
+Protagonist action timeline:
+
+0.000000–1.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000–3.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000–5.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.
+
+Exact current reference visibility (local half-open intervals):
+the camera-owner racing sedan with a seated driver (car, identity tag #f736df): visible only during [0.000000,5.166667)s.
+the protagonist (player, identity tag #a7f736): visible only during [0.000000,5.166667)s.
+The reference alone determines whether each person or object is on screen. Appearance descriptions do not add invisible subjects. Keep people seated inside their vehicles whenever the reference does so; do not render an additional person outside. Identity-tag colors and white facing marks are controls, not target clothes or surfaces. Never render boxes, wireframes, proxy colors, front marks, interface overlays or duplicate subjects.
+Only the source ends at local 3.750000 seconds in this final window; the remaining 1.416667 seconds are tail padding and will not be delivered.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/long_video/long30-022-coast-coast/follow/mixed/prompts.cwm.READY.json b/prompts/long_video/long30-022-coast-coast/follow/mixed/prompts.cwm.READY.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1e762e7b8afbfd927a07f98848751880b5ed097
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/mixed/prompts.cwm.READY.json
@@ -0,0 +1,25 @@
+{
+  "method": "cwm",
+  "sample": "long30-022-coast-coast__follow",
+  "scene_number": "022",
+  "camera": "follow",
+  "style": "realistic",
+  "steps": 20,
+  "windows": 8,
+  "window_frames": 124,
+  "stride": 90,
+  "overlap": 34,
+  "deliver": 720,
+  "fps": 24,
+  "seed": 20260909,
+  "prompts": [
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 is a diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels, the same wetsuit, and fins, seated in the driver's seat of the rival sedan whenever that driver is visible in the reference. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20132.630000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 59 metres along its route; the vehicle moves about 59 metres along its route; the vehicle moves about 60 metres along its route; the vehicle moves about 85 metres along its route; the vehicle moves about 62 metres along its route.; 2.630000\u20135.000000s: The protagonist's sedan pulls out alongside a rival sedan and tries to pass; the protagonist moves about 63 metres along its route; the vehicle moves about 63 metres along its route; the vehicle moves about 55 metres along its route; the vehicle moves about 77 metres along its route; the vehicle moves about 49 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20130.630000s: The protagonist's sedan pulls out alongside a rival sedan and tries to pass; the protagonist moves about 17 metres along its route; the vehicle moves about 17 metres along its route; the vehicle moves about 14 metres along its route; the vehicle moves about 20 metres along its route.; 0.630000\u20132.670000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 60 metres along its route; the vehicle moves about 60 metres along its route; the vehicle moves about 48 metres along its route; the vehicle moves about 64 metres along its route.; 2.670000\u20133.930000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 38 metres along its route; the vehicle moves about 38 metres along its route; the vehicle moves about 30 metres along its route; the vehicle moves about 42 metres along its route.; 3.930000\u20135.000000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 36 metres along its route; the vehicle moves about 36 metres along its route; the vehicle moves about 26 metres along its route; the vehicle moves about 36 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20130.200000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 5 metres along its route; the vehicle moves about 5 metres along its route; the vehicle moves about 4 metres along its route; the vehicle moves about 6 metres along its route.; 0.200000\u20133.570000s: The protagonist's sedan follows the rival sedan ahead along the track, holding its line; the protagonist moves about 117 metres along its route; the vehicle moves about 118 metres along its route; the vehicle moves about 84 metres along its route; the vehicle moves about 113 metres along its route.; 3.570000\u20135.000000s: The protagonist's sedan brakes for the corner and turns in; the protagonist moves about 47 metres along its route; the vehicle moves about 47 metres along its route; the vehicle moves about 35 metres along its route; the vehicle moves about 47 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20131.070000s: The protagonist's sedan brakes for the corner and turns in; the protagonist moves about 35 metres along its route; the vehicle moves about 35 metres along its route; the vehicle moves about 35 metres along its route.; 1.070000\u20133.530000s: The protagonist's sedan pulls out alongside a rival sedan and tries to pass; the protagonist moves about 95 metres along its route; the vehicle moves about 95 metres along its route; the vehicle moves about 84 metres along its route.; 3.530000\u20135.000000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 61 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 50 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 105 metres along its route; the vehicle moves about 105 metres along its route; the vehicle moves about 86 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same overtaking stretch; the protagonist moves about 102 metres along its route; the vehicle moves about 102 metres along its route; the vehicle moves about 86 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20133.400000s: The protagonist's sedan pulls out and passes a rival sedan; the protagonist moves about 136 metres along its route; the vehicle moves about 136 metres along its route; the vehicle moves about 117 metres along its route.; 3.400000\u20135.000000s: The protagonist's sedan cruises along the track at steady speed while the rival sedans keep racing on their own lines; the protagonist moves about 62 metres along its route; the vehicle moves about 62 metres along its route; the vehicle moves about 55 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 0.000000\u20132.500000s: The protagonist's sedan cruises along the track at steady speed while the rival sedans keep racing on their own lines; the protagonist moves about 98 metres along its route; the vehicle moves about 98 metres along its route; the vehicle moves about 88 metres along its route.; 2.500000\u20135.000000s: The protagonist keeps going through the same cruise stretch; the protagonist moves about 98 metres along its route; the vehicle moves about 98 metres along its route; the vehicle moves about 90 metres along its route.; 5.000000\u20135.166667s: Everyone and the camera simply keep moving as before.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n",
+    "Appearance and scene:\n\nRendering style of the whole frame: A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials.\n\nThe protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. Every described entity retains this appearance only while visible in the reference.\n\nThe setting is a coastal beach with sloping sand and large shoreline rocks. Preserve the reference spacing, masses and open routes. The player sedan, a complete graphite-grey racing sedan with a seated driver; a complete racing sedan with a seated driver. The camera follows the recorded third-person path of the conditioning reference. Early-morning low sun casts long, soft shadows across the wet sand, with a cool, desaturated colour temperature and a single warm accent hue highlighting the edges of the rocks and vehicles. An asphalt road with white lane markings and low metal roadside barriers runs through the scene exactly where the conditioning reference shows it; keep it a real paved road for the whole clip. Trees stand only where the conditioning reference shows them. \n\nProtagonist action timeline:\n\n0.000000\u20131.420000s: the already-given end of the previous clip; the action carries on from it without any restart; 1.420000\u20133.750000s: The protagonist carries on with the recorded movement of the conditioning reference; 3.750000\u20135.166667s: The recording has ended: everyone holds the final pose and position and the camera holds its final pose.\n\n\nOnly depict entities visible in the current conditioning frames; appearance text never adds an absent entity. Keep drivers inside their vehicles, without an extra person outside.\n"
+  ]
+}
\ No newline at end of file
diff --git a/prompts/long_video/long30-022-coast-coast/follow/white/prompts.c2r.json b/prompts/long_video/long30-022-coast-coast/follow/white/prompts.c2r.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e8de13026c1532f7d9f9f1f498efabc55a9da1b
--- /dev/null
+++ b/prompts/long_video/long30-022-coast-coast/follow/white/prompts.c2r.json
@@ -0,0 +1,126 @@
+[
+  {
+    "segment": 0,
+    "t0": 0.0,
+    "t1": 5.0,
+    "text": "A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 is a diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels, the same wetsuit, and fins, seated in the driver's seat of the rival sedan whenever that driver is visible in the reference. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. The protagonist's sedan pulls out alongside a rival sedan and tries to pass. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "bb5cb027982830fb412f334d7eb7442bc7889d6aa0c473a3e67cecfd2146fb96",
+    "events": [
+      {
+        "t0": 0.0,
+        "t1": 2.63,
+        "window": 0
+      },
+      {
+        "t0": 2.63,
+        "t1": 5.0,
+        "window": 0
+      },
+      {
+        "t0": 4.38,
+        "t1": 6.42,
+        "window": 1
+      }
+    ]
+  },
+  {
+    "segment": 1,
+    "t0": 5.0,
+    "t1": 10.0,
+    "text": "A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 is a diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels, the same wetsuit, and fins, seated in the driver's seat of the rival sedan whenever that driver is visible in the reference. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. The protagonist's sedan pulls out and passes a rival sedan. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "9fb659d96fcd45dceee6bb23755f32b8a0a3efac504400454de049cf33d1e833",
+    "events": [
+      {
+        "t0": 4.38,
+        "t1": 6.42,
+        "window": 1
+      },
+      {
+        "t0": 6.42,
+        "t1": 7.7,
+        "window": 1
+      },
+      {
+        "t0": 7.68,
+        "t1": 11.07,
+        "window": 1
+      }
+    ]
+  },
+  {
+    "segment": 2,
+    "t0": 10.0,
+    "t1": 15.0,
+    "text": "A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 is a diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels, the same wetsuit, and fins, seated in the driver's seat of the rival sedan whenever that driver is visible in the reference. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan follows the rival sedan ahead along the track, holding its line. The protagonist's sedan brakes for the corner and turns in. The protagonist's sedan pulls out alongside a rival sedan and tries to pass. The protagonist's sedan pulls out and passes a rival sedan. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "81dba1cac8bdb049f6d237a364e6ffa57459f4b6ecb7bb5df82b72e32e3ebfd9",
+    "events": [
+      {
+        "t0": 7.68,
+        "t1": 11.07,
+        "window": 1
+      },
+      {
+        "t0": 11.07,
+        "t1": 12.5,
+        "window": 2
+      },
+      {
+        "t0": 12.32,
+        "t1": 14.78,
+        "window": 3
+      },
+      {
+        "t0": 14.78,
+        "t1": 17.5,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 3,
+    "t0": 15.0,
+    "t1": 20.0,
+    "text": "A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 is a diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels, the same wetsuit, and fins, seated in the driver's seat of the rival sedan whenever that driver is visible in the reference. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan pulls out and passes a rival sedan. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "dab4d3bbf7c9b082e57940b93468d3b47f94e722c543c71e00b08e9b4c3dba9b",
+    "events": [
+      {
+        "t0": 14.78,
+        "t1": 22.15,
+        "window": 3
+      }
+    ]
+  },
+  {
+    "segment": 4,
+    "t0": 20.0,
+    "t1": 25.0,
+    "text": "A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 is a diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels, the same wetsuit, and fins, seated in the driver's seat of the rival sedan whenever that driver is visible in the reference. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan pulls out and passes a rival sedan. The protagonist's sedan cruises along the track at steady speed while the rival sedans keep racing on their own lines. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "fbfddf7daf265693272880d1be916cb53cb4be62e71b239712d9a21ae7a1bbcc",
+    "events": [
+      {
+        "t0": 18.75,
+        "t1": 22.15,
+        "window": 5
+      },
+      {
+        "t0": 22.15,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  },
+  {
+    "segment": 5,
+    "t0": 25.0,
+    "t1": 30.0,
+    "text": "A realistic open coastline in bright maritime light: wet sand, surf-darkened rock, sea spray and a clear horizon, with natural human anatomy, believable scale and detailed physical materials. The protagonist sits in the driver's seat inside the closed cabin of the racing sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The camera-owner racing sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The AI-driven rival sedan with a seated driver is a complete graphite-grey racing sedan with a rear wing, featuring doors, windows, and wheels. The seated driver of rival racing car 1 sits in the driver's seat inside the closed cabin of the third rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 2 sits in the driver's seat inside the closed cabin of the farther rival sedan, entirely within its bodywork, out of sight behind dark reflective glazing, and is never drawn as a separate figure outside the vehicle. The seated driver of rival racing car 3 is a diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels, the same wetsuit, and fins, seated in the driver's seat of the rival sedan whenever that driver is visible in the reference. Every described entity retains this appearance only while visible in the reference. The protagonist's sedan cruises along the track at steady speed while the rival sedans keep racing on their own lines. Only people visible in the control appear. Drivers remain inside their vehicles.",
+    "text_sha256": "89d0479cafcca862c481a92c1489511bac89fdf8808f4d8552d4ae32f54223dd",
+    "events": [
+      {
+        "t0": 22.15,
+        "t1": 25.0,
+        "window": 5
+      }
+    ]
+  }
+]
diff --git a/prompts/neon/ndm-05-cam-static_obj-enter-death/follow/colored-obb/prompt.txt b/prompts/neon/ndm-05-cam-static_obj-enter-death/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..155c398c81da513563d33858b26bc166adbe649b
--- /dev/null
+++ b/prompts/neon/ndm-05-cam-static_obj-enter-death/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from <Picture 1> and the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a tactical standoff in a paved industrial yard where a protagonist in a black vest faces three armed men near warehouses and cargo containers, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 3> (armed person 1), <Subject 4> (armed person 2), <Subject 5> (armed person 3), <Subject 6> (armed person 4); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 1.9 s.
+<Subject 1> is the protagonist: in <Picture 1> a male figure seen from behind, wearing a black tactical vest over a dark shirt, olive cargo trousers, and tan work boots. He stands in a ready stance facing the group ahead; in <Video 1> the human proxy tagged #7bf4e5, whose white face is its front.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a vast, cracked asphalt tarmac marked with mustard painted lines. The background features large corrugated metal industrial warehouses with roll-up doors, a central loading gantry, and a tall concrete observation tower. To the right, stacks of blue and rust-coloured shipping containers sit near a building. Distant grey mountain ridges rise against a pale, overcast sky; in <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is armed person 1: in <Picture 1> a man with dark hair wearing a grey hoodie, black track pants with white side stripes, and white sneakers. He stands in the middle distance holding a handgun with both hands; in <Video 1> the human proxy tagged #f47b7b, whose white face is its front.
+<Subject 4> is armed person 2: in <Picture 1> a man with dark hair wearing a sage jacket over a grey shirt, blue jeans, and white sneakers. He stands to the right, aiming a handgun forward; in <Video 1> the human proxy tagged #7baaf4, whose white face is its front.
+<Subject 5> is armed person 3: in <Picture 1> a man with dark hair wearing a maroon t-shirt, khaki trousers, and dark shoes. He stands further back to the left, holding a handgun; in <Video 1> the human proxy tagged #cdf47b that starts on the left of the frame, whose white face is its front.
+<Subject 6> is armed person 4: appearance is not established by <Picture 1>; it enters the frame at about 1.9 s; in <Video 1> the human proxy tagged #f47be9 that starts on the left of the frame, whose white face is its front; preserve its recorded identity without inferring unseen clothing or materials from the image.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> shooting armed person 4 down in an armed standoff as a newcomer walks in and the others fire back, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 6> (enters [Shot 1] at about 1.9 s): partially_preserved - retain identity, geometry, trajectory and interaction timing from <Video 1>; appearance is not established by <Picture 1>, so do not claim to preserve unseen details from that image.
+
+detailed_description:
+Photorealistic natural lighting, physically plausible scale, detailed real-world materials and natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a vast, cracked asphalt tarmac marked with mustard painted lines. The background features large corrugated metal industrial warehouses with roll-up doors, a central loading gantry, and a tall concrete observation tower. To the right, stacks of blue and rust-coloured shipping containers sit near a building. Distant grey mountain ridges rise against a pale, overcast sky, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is a male figure seen from behind, wearing a black tactical vest over a dark shirt, olive cargo trousers, and tan work boots. He stands in a ready stance facing the group ahead, as seen in <Picture 1>. <Subject 3>, armed person 1, is a man with dark hair wearing a grey hoodie, black track pants with white side stripes, and white sneakers. He stands in the middle distance holding a handgun with both hands, as seen in <Picture 1>. <Subject 4>, armed person 2, is a man with dark hair wearing a sage jacket over a grey shirt, blue jeans, and white sneakers. He stands to the right, aiming a handgun forward, as seen in <Picture 1>. <Subject 5>, armed person 3, is a man with dark hair wearing a maroon t-shirt, khaki trousers, and dark shoes. He stands further back to the left, holding a handgun, as seen in <Picture 1>. <Subject 6>, armed person 4, has no discernible appearance in <Picture 1>; preserve its recorded identity and motion from <Video 1> without treating unspecified appearance as image evidence. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:03.267, the protagonist keeps a handgun raised in both hands, aimed at the armed people in front of them, and fires at them; they fire back while another armed person walks into view; armed person 4 moves about 5 metres along its route. From 00:03.267 to 00:05.000, the protagonist shoots armed person 4, who collapses to the ground within about half a second, then lies motionless for the rest of the shot; everyone else stays on their feet, the protagonist keeps the handgun raised and the other armed people keep firing back; armed person 4 drops about half a metre. Keep the reference's fixed 51.246-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The low hum of wind across the open tarmac and the distant clatter of metal containers, punctuated by the sharp mechanical clicks of firearms being cocked.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/neon/ndm-05-cam-static_obj-enter-death/follow/white/prompt.txt b/prompts/neon/ndm-05-cam-static_obj-enter-death/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..72aabdce65f521e6e422399ccbc3c2834df5dace
--- /dev/null
+++ b/prompts/neon/ndm-05-cam-static_obj-enter-death/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from <Picture 1> and the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a tactical standoff in a paved industrial yard where a protagonist in a black vest faces three armed men near warehouses and cargo containers, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 3> (armed person 1), <Subject 4> (armed person 2), <Subject 5> (armed person 3), <Subject 6> (armed person 4); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 1.9 s.
+<Subject 1> is the protagonist: in <Picture 1> a male figure seen from behind, wearing a black tactical vest over a dark shirt, olive cargo trousers, and tan work boots. He stands in a ready stance facing the group ahead; in <Video 1> the grey block-figure proxy.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a vast, cracked asphalt tarmac marked with mustard painted lines. The background features large corrugated metal industrial warehouses with roll-up doors, a central loading gantry, and a tall concrete observation tower. To the right, stacks of blue and rust-coloured shipping containers sit near a building. Distant grey mountain ridges rise against a pale, overcast sky; in <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is armed person 1: in <Picture 1> a man with dark hair wearing a grey hoodie, black track pants with white side stripes, and white sneakers. He stands in the middle distance holding a handgun with both hands; in <Video 1> the grey block-figure proxy.
+<Subject 4> is armed person 2: in <Picture 1> a man with dark hair wearing a sage jacket over a grey shirt, blue jeans, and white sneakers. He stands to the right, aiming a handgun forward; in <Video 1> the grey block-figure proxy.
+<Subject 5> is armed person 3: in <Picture 1> a man with dark hair wearing a maroon t-shirt, khaki trousers, and dark shoes. He stands further back to the left, holding a handgun; in <Video 1> the grey block-figure proxy that starts on the left of the frame.
+<Subject 6> is armed person 4: appearance is not established by <Picture 1>; it enters the frame at about 1.9 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; preserve its recorded identity without inferring unseen clothing or materials from the image.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> shooting armed person 4 down in an armed standoff as a newcomer walks in and the others fire back, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (enters [Shot 1] at about 1.9 s): partially_preserved - retain identity, geometry, trajectory and interaction timing from <Video 1>; appearance is not established by <Picture 1>, so do not claim to preserve unseen details from that image.
+
+detailed_description:
+Photorealistic natural lighting, physically plausible scale, detailed real-world materials and natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a vast, cracked asphalt tarmac marked with mustard painted lines. The background features large corrugated metal industrial warehouses with roll-up doors, a central loading gantry, and a tall concrete observation tower. To the right, stacks of blue and rust-coloured shipping containers sit near a building. Distant grey mountain ridges rise against a pale, overcast sky, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is a male figure seen from behind, wearing a black tactical vest over a dark shirt, olive cargo trousers, and tan work boots. He stands in a ready stance facing the group ahead, as seen in <Picture 1>. <Subject 3>, armed person 1, is a man with dark hair wearing a grey hoodie, black track pants with white side stripes, and white sneakers. He stands in the middle distance holding a handgun with both hands, as seen in <Picture 1>. <Subject 4>, armed person 2, is a man with dark hair wearing a sage jacket over a grey shirt, blue jeans, and white sneakers. He stands to the right, aiming a handgun forward, as seen in <Picture 1>. <Subject 5>, armed person 3, is a man with dark hair wearing a maroon t-shirt, khaki trousers, and dark shoes. He stands further back to the left, holding a handgun, as seen in <Picture 1>. <Subject 6>, armed person 4, has no discernible appearance in <Picture 1>; preserve its recorded identity and motion from <Video 1> without treating unspecified appearance as image evidence. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:03.267, the protagonist keeps a handgun raised in both hands, aimed at the armed people in front of them, and fires at them; they fire back while another armed person walks into view; armed person 4 moves about 5 metres along its route. From 00:03.267 to 00:05.000, the protagonist shoots armed person 4, who collapses to the ground within about half a second, then lies motionless for the rest of the shot; everyone else stays on their feet, the protagonist keeps the handgun raised and the other armed people keep firing back; armed person 4 drops about half a metre. Keep the reference's fixed 51.246-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The low hum of wind across the open tarmac and the distant clatter of metal containers, punctuated by the sharp mechanical clicks of firearms being cocked.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/colored-obb/prompt.txt b/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..64bd945b33d5df76890298358f2af80995864b8e
--- /dev/null
+++ b/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from <Picture 1> and the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a standoff in a paved industrial yard where a central protagonist in tactical gear faces two armed men near warehouses and cargo containers, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 1> (the protagonist), <Subject 3> (armed person 1), <Subject 4> (armed person 2); nothing else in <Video 1> is a person or an interaction object. The proxy tagged #cdf47b (person) never enters the frame of this shot: no figure or object appears for it and none must be added.
+<Subject 1> is the protagonist: in <Picture 1> a bald adult male seen from behind, wearing a grey tactical bulletproof vest over a black long-sleeved shirt, camouflage cargo trousers, and black combat boots. He is holding a handgun with both hands, aiming forward; in <Video 1> the human proxy tagged #f47be9, whose white face is its front.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a vast paved asphalt lot marked with mustard painted lines and cross-hatching. The background features large industrial warehouses with corrugated metal siding and roll-up doors. A central loading gantry structure stands between buildings. To the right, stacked shipping containers in forest green and rust red are visible. Distant mountain ridges fade into a hazy, overcast sky; in <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is armed person 1: in <Picture 1> a man with short brown hair, wearing a brown leather jacket, khaki trousers, and dark shoes. He is holding a handgun with both hands, aiming towards the protagonist; in <Video 1> the human proxy tagged #f47b7b that starts on the left of the frame, whose white face is its front.
+<Subject 4> is armed person 2: in <Picture 1> a man with dark hair, wearing a navy blazer over a dark shirt, black trousers, and black sneakers. He is holding a handgun with both hands, aiming towards the protagonist; in <Video 1> the human proxy tagged #7baaf4, whose white face is its front.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> shooting armed person 1 down while armed people on the move fire back, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic natural lighting, physically plausible scale, detailed real-world materials and natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a vast paved asphalt lot marked with mustard painted lines and cross-hatching. The background features large industrial warehouses with corrugated metal siding and roll-up doors. A central loading gantry structure stands between buildings. To the right, stacked shipping containers in forest green and rust red are visible. Distant mountain ridges fade into a hazy, overcast sky, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is a bald adult male seen from behind, wearing a grey tactical bulletproof vest over a black long-sleeved shirt, camouflage cargo trousers, and black combat boots. He is holding a handgun with both hands, aiming forward, as seen in <Picture 1>. <Subject 3>, armed person 1, is a man with short brown hair, wearing a brown leather jacket, khaki trousers, and dark shoes. He is holding a handgun with both hands, aiming towards the protagonist, as seen in <Picture 1>. <Subject 4>, armed person 2, is a man with dark hair, wearing a navy blazer over a dark shirt, black trousers, and black sneakers. He is holding a handgun with both hands, aiming towards the protagonist, as seen in <Picture 1>. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.767, the protagonist keeps a handgun raised in both hands, aimed at the armed people in front of them, and fires at them; some of them move slowly across the open ground and fire back; armed person 2 moves about 2 metres along its route; the protagonist moves about 4 metres along its route. From 00:02.767 to 00:05.000, the protagonist shoots armed person 1, who collapses to the ground within about half a second, then lies motionless for the rest of the shot; everyone else stays on their feet, the protagonist keeps the handgun raised and the other armed people keep firing back; armed person 1 drops about half a metre; armed person 2 moves about 2 metres along its route; the protagonist moves about 4 metres along its route. Keep the reference's fixed 50.984-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The ambient hum of wind across the open yard and distant traffic, punctuated by the sharp metallic click of a pistol slide being racked.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/mixed/prompt.txt b/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..75aa67f96dd1fff978d3bb33da9c50d936df304e
--- /dev/null
+++ b/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from <Picture 1> and the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a standoff in a paved industrial yard where a central protagonist in tactical gear faces two armed men near warehouses and cargo containers, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 3> (armed person 1), <Subject 4> (armed person 2); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist: in <Picture 1> a bald adult male seen from behind, wearing a grey tactical bulletproof vest over a black long-sleeved shirt, camouflage cargo trousers, and black combat boots. He is holding a handgun with both hands, aiming forward; in <Video 1> the blue mannequin proxy.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a vast paved asphalt lot marked with mustard painted lines and cross-hatching. The background features large industrial warehouses with corrugated metal siding and roll-up doors. A central loading gantry structure stands between buildings. To the right, stacked shipping containers in forest green and rust red are visible. Distant mountain ridges fade into a hazy, overcast sky; in <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is armed person 1: in <Picture 1> a man with short brown hair, wearing a brown leather jacket, khaki trousers, and dark shoes. He is holding a handgun with both hands, aiming towards the protagonist; in <Video 1> the blue mannequin proxy that starts on the left of the frame.
+<Subject 4> is armed person 2: in <Picture 1> a man with dark hair, wearing a navy blazer over a dark shirt, black trousers, and black sneakers. He is holding a handgun with both hands, aiming towards the protagonist; in <Video 1> the blue mannequin proxy.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> shooting armed person 1 down while armed people on the move fire back, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic natural lighting, physically plausible scale, detailed real-world materials and natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a vast paved asphalt lot marked with mustard painted lines and cross-hatching. The background features large industrial warehouses with corrugated metal siding and roll-up doors. A central loading gantry structure stands between buildings. To the right, stacked shipping containers in forest green and rust red are visible. Distant mountain ridges fade into a hazy, overcast sky, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is a bald adult male seen from behind, wearing a grey tactical bulletproof vest over a black long-sleeved shirt, camouflage cargo trousers, and black combat boots. He is holding a handgun with both hands, aiming forward, as seen in <Picture 1>. <Subject 3>, armed person 1, is a man with short brown hair, wearing a brown leather jacket, khaki trousers, and dark shoes. He is holding a handgun with both hands, aiming towards the protagonist, as seen in <Picture 1>. <Subject 4>, armed person 2, is a man with dark hair, wearing a navy blazer over a dark shirt, black trousers, and black sneakers. He is holding a handgun with both hands, aiming towards the protagonist, as seen in <Picture 1>. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.767, the protagonist keeps a handgun raised in both hands, aimed at the armed people in front of them, and fires at them; some of them move slowly across the open ground and fire back; armed person 2 moves about 2 metres along its route; the protagonist moves about 4 metres along its route. From 00:02.767 to 00:05.000, the protagonist shoots armed person 1, who collapses to the ground within about half a second, then lies motionless for the rest of the shot; everyone else stays on their feet, the protagonist keeps the handgun raised and the other armed people keep firing back; armed person 1 drops about half a metre; armed person 2 moves about 2 metres along its route; the protagonist moves about 4 metres along its route. Keep the reference's fixed 50.984-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The ambient hum of wind across the open yard and distant traffic, punctuated by the sharp metallic click of a pistol slide being racked.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/white/prompt.txt b/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b463c2351694cbf2ea9544e86a1ea662853f6544
--- /dev/null
+++ b/prompts/neon/ndm-08-cam-dynamic_obj-dynamic-death/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from <Picture 1> and the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a standoff in a paved industrial yard where a central protagonist in tactical gear faces two armed men near warehouses and cargo containers, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 3> (armed person 1), <Subject 4> (armed person 2); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist: in <Picture 1> a bald adult male seen from behind, wearing a grey tactical bulletproof vest over a black long-sleeved shirt, camouflage cargo trousers, and black combat boots. He is holding a handgun with both hands, aiming forward; in <Video 1> the grey block-figure proxy.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a vast paved asphalt lot marked with mustard painted lines and cross-hatching. The background features large industrial warehouses with corrugated metal siding and roll-up doors. A central loading gantry structure stands between buildings. To the right, stacked shipping containers in forest green and rust red are visible. Distant mountain ridges fade into a hazy, overcast sky; in <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is armed person 1: in <Picture 1> a man with short brown hair, wearing a brown leather jacket, khaki trousers, and dark shoes. He is holding a handgun with both hands, aiming towards the protagonist; in <Video 1> the grey block-figure proxy that starts on the left of the frame.
+<Subject 4> is armed person 2: in <Picture 1> a man with dark hair, wearing a navy blazer over a dark shirt, black trousers, and black sneakers. He is holding a handgun with both hands, aiming towards the protagonist; in <Video 1> the grey block-figure proxy.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> shooting armed person 1 down while armed people on the move fire back, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic natural lighting, physically plausible scale, detailed real-world materials and natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a vast paved asphalt lot marked with mustard painted lines and cross-hatching. The background features large industrial warehouses with corrugated metal siding and roll-up doors. A central loading gantry structure stands between buildings. To the right, stacked shipping containers in forest green and rust red are visible. Distant mountain ridges fade into a hazy, overcast sky, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is a bald adult male seen from behind, wearing a grey tactical bulletproof vest over a black long-sleeved shirt, camouflage cargo trousers, and black combat boots. He is holding a handgun with both hands, aiming forward, as seen in <Picture 1>. <Subject 3>, armed person 1, is a man with short brown hair, wearing a brown leather jacket, khaki trousers, and dark shoes. He is holding a handgun with both hands, aiming towards the protagonist, as seen in <Picture 1>. <Subject 4>, armed person 2, is a man with dark hair, wearing a navy blazer over a dark shirt, black trousers, and black sneakers. He is holding a handgun with both hands, aiming towards the protagonist, as seen in <Picture 1>. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.767, the protagonist keeps a handgun raised in both hands, aimed at the armed people in front of them, and fires at them; some of them move slowly across the open ground and fire back; armed person 2 moves about 2 metres along its route; the protagonist moves about 4 metres along its route. From 00:02.767 to 00:05.000, the protagonist shoots armed person 1, who collapses to the ground within about half a second, then lies motionless for the rest of the shot; everyone else stays on their feet, the protagonist keeps the handgun raised and the other armed people keep firing back; armed person 1 drops about half a metre; armed person 2 moves about 2 metres along its route; the protagonist moves about 4 metres along its route. Keep the reference's fixed 50.984-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The ambient hum of wind across the open yard and distant traffic, punctuated by the sharp metallic click of a pistol slide being racked.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/colored-obb/prompt.txt b/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9004d00107c8008d6753e0d2a7cdb73f0fdf05a1
--- /dev/null
+++ b/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/colored-obb/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from <Picture 1> and the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a protagonist stands on a concrete airport apron facing two armed adversaries in front of large industrial hangars under a sunset sky, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 3> (armed person 1), <Subject 4> (armed person 2), <Subject 5> (armed person 3); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 1.9 s.
+<Subject 1> is the protagonist: in <Picture 1> a man with a shaved head wearing a black tactical vest over a dark shirt, olive cargo trousers and tan work boots, holding a handgun with both hands in a firing stance; in <Video 1> the human proxy tagged #f47be9, whose white face is its front.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a wide paved airport apron with yellow taxiway markings and visible cracks, flanked by large corrugated metal hangars with segmented doors and small windows. A tall floodlight mast stands between the buildings, and distant mountain ridges are visible under a soft sunset sky; in <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is armed person 1: in <Picture 1> a man with dark hair wearing a forest bomber jacket, a white t-shirt, blue jeans and white sneakers, holding a handgun with both hands in a firing stance; in <Video 1> the human proxy tagged #f47b7b that starts on the left of the frame, whose white face is its front.
+<Subject 4> is armed person 2: in <Picture 1> a man with a shaved head wearing a grey hoodie, black trousers and grey sneakers, holding a handgun with both hands in a firing stance; in <Video 1> the human proxy tagged #7baaf4, whose white face is its front.
+<Subject 5> is armed person 3: appearance is not established by <Picture 1>; it enters the frame at about 1.9 s; in <Video 1> the human proxy tagged #cdf47b that starts on the left of the frame, whose white face is its front; preserve its recorded identity without inferring unseen clothing or materials from the image.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> shooting armed person 3 down in an armed standoff as a newcomer walks in and the others fire back, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 5> (enters [Shot 1] at about 1.9 s): partially_preserved - retain identity, geometry, trajectory and interaction timing from <Video 1>; appearance is not established by <Picture 1>, so do not claim to preserve unseen details from that image.
+
+detailed_description:
+Photorealistic natural lighting, physically plausible scale, detailed real-world materials and natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a wide paved airport apron with yellow taxiway markings and visible cracks, flanked by large corrugated metal hangars with segmented doors and small windows. A tall floodlight mast stands between the buildings, and distant mountain ridges are visible under a soft sunset sky, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is a man with a shaved head wearing a black tactical vest over a dark shirt, olive cargo trousers and tan work boots, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 3>, armed person 1, is a man with dark hair wearing a forest bomber jacket, a white t-shirt, blue jeans and white sneakers, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 4>, armed person 2, is a man with a shaved head wearing a grey hoodie, black trousers and grey sneakers, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 5>, armed person 3, has no discernible appearance in <Picture 1>; preserve its recorded identity and motion from <Video 1> without treating unspecified appearance as image evidence. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:03.267, the protagonist keeps a handgun raised in both hands, aimed at the armed people in front of them, and fires at them; they fire back while another armed person walks into view; armed person 3 moves about 5 metres along its route. From 00:03.267 to 00:05.000, the protagonist shoots armed person 3, who collapses to the ground within about half a second, then lies motionless for the rest of the shot; everyone else stays on their feet, the protagonist keeps the handgun raised and the other armed people keep firing back; armed person 3 drops about half a metre. Keep the reference's fixed 51.784-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The ambient hum of distant airport operations and wind, punctuated by the sharp crack of gunfire and the scuff of boots on concrete.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/mixed/prompt.txt b/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..96972b58f9170556364c603049d3d0f9a1d275c6
--- /dev/null
+++ b/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/mixed/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from <Picture 1> and the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a protagonist stands on a concrete airport apron facing two armed adversaries in front of large industrial hangars under a sunset sky, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 3> (armed person 1), <Subject 4> (armed person 2), <Subject 5> (armed person 3); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 1.9 s.
+<Subject 1> is the protagonist: in <Picture 1> a man with a shaved head wearing a black tactical vest over a dark shirt, olive cargo trousers and tan work boots, holding a handgun with both hands in a firing stance; in <Video 1> the blue mannequin proxy.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a wide paved airport apron with yellow taxiway markings and visible cracks, flanked by large corrugated metal hangars with segmented doors and small windows. A tall floodlight mast stands between the buildings, and distant mountain ridges are visible under a soft sunset sky; in <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is armed person 1: in <Picture 1> a man with dark hair wearing a forest bomber jacket, a white t-shirt, blue jeans and white sneakers, holding a handgun with both hands in a firing stance; in <Video 1> the blue mannequin proxy that starts on the left of the frame.
+<Subject 4> is armed person 2: in <Picture 1> a man with a shaved head wearing a grey hoodie, black trousers and grey sneakers, holding a handgun with both hands in a firing stance; in <Video 1> the blue mannequin proxy.
+<Subject 5> is armed person 3: appearance is not established by <Picture 1>; it enters the frame at about 1.9 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; preserve its recorded identity without inferring unseen clothing or materials from the image.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> shooting armed person 3 down in an armed standoff as a newcomer walks in and the others fire back, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 5> (enters [Shot 1] at about 1.9 s): partially_preserved - retain identity, geometry, trajectory and interaction timing from <Video 1>; appearance is not established by <Picture 1>, so do not claim to preserve unseen details from that image.
+
+detailed_description:
+Photorealistic natural lighting, physically plausible scale, detailed real-world materials and natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a wide paved airport apron with yellow taxiway markings and visible cracks, flanked by large corrugated metal hangars with segmented doors and small windows. A tall floodlight mast stands between the buildings, and distant mountain ridges are visible under a soft sunset sky, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is a man with a shaved head wearing a black tactical vest over a dark shirt, olive cargo trousers and tan work boots, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 3>, armed person 1, is a man with dark hair wearing a forest bomber jacket, a white t-shirt, blue jeans and white sneakers, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 4>, armed person 2, is a man with a shaved head wearing a grey hoodie, black trousers and grey sneakers, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 5>, armed person 3, has no discernible appearance in <Picture 1>; preserve its recorded identity and motion from <Video 1> without treating unspecified appearance as image evidence. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:03.267, the protagonist keeps a handgun raised in both hands, aimed at the armed people in front of them, and fires at them; they fire back while another armed person walks into view; armed person 3 moves about 5 metres along its route. From 00:03.267 to 00:05.000, the protagonist shoots armed person 3, who collapses to the ground within about half a second, then lies motionless for the rest of the shot; everyone else stays on their feet, the protagonist keeps the handgun raised and the other armed people keep firing back; armed person 3 drops about half a metre. Keep the reference's fixed 51.784-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The ambient hum of distant airport operations and wind, punctuated by the sharp crack of gunfire and the scuff of boots on concrete.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/white/prompt.txt b/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..251b1576eaa8c6aaa34a79c0ae4cdd9c396c4dec
--- /dev/null
+++ b/prompts/neon/ndm-09-cam-static_obj-enter-death/follow/white/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from <Picture 1> and the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a protagonist stands on a concrete airport apron facing two armed adversaries in front of large industrial hangars under a sunset sky, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 3> (armed person 1), <Subject 4> (armed person 2), <Subject 5> (armed person 3); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 1.9 s.
+<Subject 1> is the protagonist: in <Picture 1> a man with a shaved head wearing a black tactical vest over a dark shirt, olive cargo trousers and tan work boots, holding a handgun with both hands in a firing stance; in <Video 1> the grey block-figure proxy.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a wide paved airport apron with yellow taxiway markings and visible cracks, flanked by large corrugated metal hangars with segmented doors and small windows. A tall floodlight mast stands between the buildings, and distant mountain ridges are visible under a soft sunset sky; in <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is armed person 1: in <Picture 1> a man with dark hair wearing a forest bomber jacket, a white t-shirt, blue jeans and white sneakers, holding a handgun with both hands in a firing stance; in <Video 1> the grey block-figure proxy that starts on the left of the frame.
+<Subject 4> is armed person 2: in <Picture 1> a man with a shaved head wearing a grey hoodie, black trousers and grey sneakers, holding a handgun with both hands in a firing stance; in <Video 1> the grey block-figure proxy.
+<Subject 5> is armed person 3: appearance is not established by <Picture 1>; it enters the frame at about 1.9 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; preserve its recorded identity without inferring unseen clothing or materials from the image.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> shooting armed person 3 down in an armed standoff as a newcomer walks in and the others fire back, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (enters [Shot 1] at about 1.9 s): partially_preserved - retain identity, geometry, trajectory and interaction timing from <Video 1>; appearance is not established by <Picture 1>, so do not claim to preserve unseen details from that image.
+
+detailed_description:
+Photorealistic natural lighting, physically plausible scale, detailed real-world materials and natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a wide paved airport apron with yellow taxiway markings and visible cracks, flanked by large corrugated metal hangars with segmented doors and small windows. A tall floodlight mast stands between the buildings, and distant mountain ridges are visible under a soft sunset sky, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is a man with a shaved head wearing a black tactical vest over a dark shirt, olive cargo trousers and tan work boots, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 3>, armed person 1, is a man with dark hair wearing a forest bomber jacket, a white t-shirt, blue jeans and white sneakers, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 4>, armed person 2, is a man with a shaved head wearing a grey hoodie, black trousers and grey sneakers, holding a handgun with both hands in a firing stance, as seen in <Picture 1>. <Subject 5>, armed person 3, has no discernible appearance in <Picture 1>; preserve its recorded identity and motion from <Video 1> without treating unspecified appearance as image evidence. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:03.267, the protagonist keeps a handgun raised in both hands, aimed at the armed people in front of them, and fires at them; they fire back while another armed person walks into view; armed person 3 moves about 5 metres along its route. From 00:03.267 to 00:05.000, the protagonist shoots armed person 3, who collapses to the ground within about half a second, then lies motionless for the rest of the shot; everyone else stays on their feet, the protagonist keeps the handgun raised and the other armed people keep firing back; armed person 3 drops about half a metre. Keep the reference's fixed 51.784-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The ambient hum of distant airport operations and wind, punctuated by the sharp crack of gunfire and the scuff of boots on concrete.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ea6aeb423179bf75accdf578b0a28cc3644a1bce
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a deep burgundy body, complete with doors, windows, and wheels, is positioned in the middle centre, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. A vintage sedan with a forest green body, complete with doors, windows, and wheels, is positioned on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #f79e36; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is advancing along a street while pedestrians and vehicles move through it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a deep burgundy body, complete with doors, windows, and wheels, is positioned in the middle centre, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. A vintage sedan with a forest green body, complete with doors, windows, and wheels, is positioned on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same palette-knife oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, stands on the middle far left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, stands on the middle right, seen only from behind, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, stands on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the street while the pedestrians and vehicles move through it; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 10 metres along its route; the vehicle moves about 9 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the traffic keeps moving; the vehicle moves about half a metre along its route; the vehicle moves about half a metre along its route. <Subject 3> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 9 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 0.8 s to 1.6 s; the passenger car in the foreground covers independently moving adult participant 1 from about 1.4 s to 1.8 s; the passenger car in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d07948434d184678d62edeaa091efadea9667c48
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a deep burgundy body, complete with doors, windows, and wheels, is positioned in the middle centre, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. A vintage sedan with a forest green body, complete with doors, windows, and wheels, is positioned on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is advancing along a street while pedestrians and vehicles move through it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a deep burgundy body, complete with doors, windows, and wheels, is positioned in the middle centre, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. A vintage sedan with a forest green body, complete with doors, windows, and wheels, is positioned on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same palette-knife oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, stands on the middle far left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, stands on the middle right, seen only from behind, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, stands on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the street while the pedestrians and vehicles move through it; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 10 metres along its route; the vehicle moves about 9 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the traffic keeps moving; the vehicle moves about half a metre along its route; the vehicle moves about half a metre along its route. <Subject 3> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 9 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 0.8 s to 1.6 s; the passenger car in the foreground covers independently moving adult participant 1 from about 1.4 s to 1.8 s; the passenger car in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..28f1d000936971a1db083dbf1bdd0884237291ee
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-003-city-traffic/first-person/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a deep burgundy body, complete with doors, windows, and wheels, is positioned in the middle centre, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. A vintage sedan with a forest green body, complete with doors, windows, and wheels, is positioned on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is advancing along a street while pedestrians and vehicles move through it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a deep burgundy body, complete with doors, windows, and wheels, is positioned in the middle centre, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. A vintage sedan with a forest green body, complete with doors, windows, and wheels, is positioned on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same palette-knife oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, stands on the middle far left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, stands on the middle right, seen only from behind, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, stands on the middle left, painted in the same palette-knife oil painting manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the street while the pedestrians and vehicles move through it; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 10 metres along its route; the vehicle moves about 9 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the traffic keeps moving; the vehicle moves about half a metre along its route; the vehicle moves about half a metre along its route. <Subject 3> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 9 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 0.8 s to 1.6 s; the passenger car in the foreground covers independently moving adult participant 1 from about 1.4 s to 1.8 s; the passenger car in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-004-city-stairs/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-004-city-stairs/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d1382ce7d1a8f2e461af8781cadb937c53ab724a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-004-city-stairs/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A dark grey pickup truck, positioned on the middle right, its metallic surface reflecting the ambient blue light and amber streetlamps, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #36f796 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36aff7 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the human proxy tagged #e7f736 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the human proxy tagged #ce36f7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A dark grey pickup truck, positioned on the middle right, its metallic surface reflecting the ambient blue light and amber streetlamps, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult male office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing on the middle left, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult male bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing on the middle right, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is an adult female student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the upper centre background, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 5 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 8 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The passenger car is one continuous object throughout and ends the clip about 13 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 3.7 s to 4.5 s; the platform-rail--1 in the foreground covers independently moving adult participant 3 from about 2.9 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-004-city-stairs/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-004-city-stairs/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..98b02564f10ce1221b63a7e51142a8239038c8c9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-004-city-stairs/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A dark grey pickup truck, positioned on the middle right, its metallic surface reflecting the ambient blue light and amber streetlamps, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A dark grey pickup truck, positioned on the middle right, its metallic surface reflecting the ambient blue light and amber streetlamps, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult male office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing on the middle left, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult male bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing on the middle right, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is an adult female student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the upper centre background, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 5 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 8 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The passenger car is one continuous object throughout and ends the clip about 13 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 3.7 s to 4.5 s; the platform-rail--1 in the foreground covers independently moving adult participant 3 from about 2.9 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-004-city-stairs/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-004-city-stairs/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..81ed86c047a74d4a8039ffc6f38d4cc4fece9088
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-004-city-stairs/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A dark grey pickup truck, positioned on the middle right, its metallic surface reflecting the ambient blue light and amber streetlamps, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A dark grey pickup truck, positioned on the middle right, its metallic surface reflecting the ambient blue light and amber streetlamps, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult male office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing on the middle left, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult male bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing on the middle right, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is an adult female student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the upper centre background, drawn in the same design-marker render sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 5 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 8 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The passenger car is one continuous object throughout and ends the clip about 13 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 3.7 s to 4.5 s; the platform-rail--1 in the foreground covers independently moving adult participant 3 from about 2.9 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a high-key pastel palette, soft diffuse studio-like light, gentle wind moving loose fabric and foliage, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4dac8a86e455dda1c369f5d46d4ad57f1994e35a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/colored-obb/prompt.txt
@@ -0,0 +1,34 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains no human proxy; nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing drawn in the same fine crosshatched etching-style drawing manner, with clearly readable features, hair and outfit; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f14a9616d77c0e80fa778e085b6a165158d56b64
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/mixed/prompt.txt
@@ -0,0 +1,34 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains no human proxy; nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing drawn in the same fine crosshatched etching-style drawing manner, with clearly readable features, hair and outfit; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..dc84bfce539d2387269841ab2a514c89164e585e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-007-city-stairsDown/first-person/white/prompt.txt
@@ -0,0 +1,34 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains no human proxy; nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing drawn in the same fine crosshatched etching-style drawing manner, with clearly readable features, hair and outfit; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-011-city-handoff/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-011-city-handoff/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ace4e9f722ecdf7433da5a9b224730458f6127ba
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-011-city-handoff/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.7 s; <Subject 8> is not in frame at 00:00.000 and enters at about 0.7 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A metal wheeled handcart with a rusted frame and wooden handles, positioned on the far left edge of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip its front toward the camera. A wooden cargo crate with visible grain and metal corner brackets, held by the protagonist near the center, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the object proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.7 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (enters [Shot 1] at about 0.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A metal wheeled handcart with a rusted frame and wooden handles, positioned on the far left edge of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip its front toward the camera. A wooden cargo crate with visible grain and metal corner brackets, held by the protagonist near the center, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs). <Subject 1>, the protagonist, has the following target appearance: A courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle ground facing slightly right, seen only from behind, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing near the center facing the player, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing on the left side of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing further back in the middle-left distance, seen only from behind, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 0.7 s; from then on it is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, currently outside the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. Independently moving adult participant 1 in the foreground covers independently moving adult participant 2 from about 4.5 s to 4.7 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 3.2 s to 4.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-011-city-handoff/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-011-city-handoff/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f22bae723841d51449d7eb601c78fb454b58b440
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-011-city-handoff/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.7 s; <Subject 8> is not in frame at 00:00.000 and enters at about 0.7 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A metal wheeled handcart with a rusted frame and wooden handles, positioned on the far left edge of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip its front toward the camera. A wooden cargo crate with visible grain and metal corner brackets, held by the protagonist near the center, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the black solid cart proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.7 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (enters [Shot 1] at about 0.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A metal wheeled handcart with a rusted frame and wooden handles, positioned on the far left edge of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip its front toward the camera. A wooden cargo crate with visible grain and metal corner brackets, held by the protagonist near the center, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs). <Subject 1>, the protagonist, has the following target appearance: A courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle ground facing slightly right, seen only from behind, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing near the center facing the player, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing on the left side of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing further back in the middle-left distance, seen only from behind, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 0.7 s; from then on it is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, currently outside the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. Independently moving adult participant 1 in the foreground covers independently moving adult participant 2 from about 4.5 s to 4.7 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 3.2 s to 4.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-011-city-handoff/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-011-city-handoff/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d535033472032ea93a93b5a7119712f91c5bcc61
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-011-city-handoff/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.7 s; <Subject 8> is not in frame at 00:00.000 and enters at about 0.7 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A metal wheeled handcart with a rusted frame and wooden handles, positioned on the far left edge of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip its front toward the camera. A wooden cargo crate with visible grain and metal corner brackets, held by the protagonist near the center, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.7 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (enters [Shot 1] at about 0.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A metal wheeled handcart with a rusted frame and wooden handles, positioned on the far left edge of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip its front toward the camera. A wooden cargo crate with visible grain and metal corner brackets, held by the protagonist near the center, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs). <Subject 1>, the protagonist, has the following target appearance: A courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle ground facing slightly right, seen only from behind, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing near the center facing the player, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing on the left side of the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing further back in the middle-left distance, seen only from behind, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 0.7 s; from then on it is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, currently outside the frame, drawn as an anime character in the same ink-brush-lined anime manner (character design: tall slender figures with small faces and long limbs), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. Independently moving adult participant 1 in the foreground covers independently moving adult participant 2 from about 4.5 s to 4.7 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 3.2 s to 4.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — ink-brush anime look: calligraphic brush outlines with tapering strokes, flat muted colours, rough paper texture and sparse painted backgrounds; a moody deep-shadow palette with luminous highlights, harsh midday sun with crisp shadows, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-016-city-handoff/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-016-city-handoff/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8c831e2ae8cffc9ccde0a063dbe5379f955d3eb8
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-016-city-handoff/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist. A rugged cargo utility pickup truck with a slate-grey body, visible doors, windows, and wheels, positioned on the left - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the cargo utility truck; in <Video 1> the object proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist. A rugged cargo utility pickup truck with a slate-grey body, visible doors, windows, and wheels, positioned on the left - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the foreground; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the cargo utility truck, realized with the materials above. <Subject 5> (the person receiving the object) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing opposite the protagonist, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the middle distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a construction worker with short black hair, wearing an amber high-visibility vest, khaki work trousers, and tan work boots, standing further back on the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a site engineer with short wavy dark hair, wearing a cream polo under a navy fleece, slate trousers, and brown boots, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 7 metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 4.4 s to 4.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-016-city-handoff/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-016-city-handoff/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8dc2e2ed27046819d3060c7fbfb28b3e9e385e88
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-016-city-handoff/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist. A rugged cargo utility pickup truck with a slate-grey body, visible doors, windows, and wheels, positioned on the left - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the cargo utility truck; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist. A rugged cargo utility pickup truck with a slate-grey body, visible doors, windows, and wheels, positioned on the left - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the foreground; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the cargo utility truck, realized with the materials above. <Subject 5> (the person receiving the object) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing opposite the protagonist, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the middle distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a construction worker with short black hair, wearing an amber high-visibility vest, khaki work trousers, and tan work boots, standing further back on the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a site engineer with short wavy dark hair, wearing a cream polo under a navy fleece, slate trousers, and brown boots, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 7 metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 4.4 s to 4.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-016-city-handoff/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-016-city-handoff/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3e0435d8cdc73984762e11d7808515fda595bbbc
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-016-city-handoff/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist. A rugged cargo utility pickup truck with a slate-grey body, visible doors, windows, and wheels, positioned on the left - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the cargo utility truck; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist. A rugged cargo utility pickup truck with a slate-grey body, visible doors, windows, and wheels, positioned on the left - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the foreground; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the cargo utility truck, realized with the materials above. <Subject 5> (the person receiving the object) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing opposite the protagonist, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the middle distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a construction worker with short black hair, wearing an amber high-visibility vest, khaki work trousers, and tan work boots, standing further back on the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a site engineer with short wavy dark hair, wearing a cream polo under a navy fleece, slate trousers, and brown boots, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 7 metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 4.4 s to 4.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-017-city-combat/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-017-city-combat/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..af9f7665487a016fadb1e123db205dfc130e7d05
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-017-city-combat/follow/colored-obb/prompt.txt
@@ -0,0 +1,64 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 human proxies standing for <Subject 1> (the protagonist), <Subject 13> (independently moving adult participant 1), <Subject 14> (independently moving adult participant 2), <Subject 15> (independently moving adult participant 3), <Subject 16> (the training opponent moving laterally and returning fire), <Subject 17> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 11> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 10> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 12> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered surface, positioned in the middle of the street. A training rifle with a matte black finish, held by the player. A vintage sedan with a deep navy body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 10> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 11> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 12> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 13> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 14> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 15> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 16> is the training opponent moving laterally and returning fire; in <Video 1> the human proxy tagged #f73677; use the target appearance described below.
+<Subject 17> is the second training opponent returning fire from the opposite side; in <Video 1> the human proxy tagged #36aff7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 17>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 11> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 12> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 17> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered surface, positioned in the middle of the street. A training rifle with a matte black finish, held by the player. A vintage sedan with a deep navy body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the short-lived shot tracer, realized with the materials above. <Subject 11> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 12> is the short-lived shot tracer, realized with the materials above. <Subject 13> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 15> (independently moving adult participant 3) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the training opponent moving laterally and returning fire) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 17> (the second training opponent returning fire from the opposite side) is a skater with shaggy chestnut hair, wearing a charcoal graphic tee, olive baggy jeans, and scuffed black skate shoes, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about one and a half metres along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 11> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and comes to rest about 12 metres from the protagonist. <Subject 11> is one continuous object throughout and comes to rest about 14 metres from the protagonist. The short-lived shot tracer in the foreground covers independently moving adult participant 1 from about 2.3 s to 3.1 s; the protagonist in the foreground covers independently moving adult participant 2 from about 1.7 s to 1.9 s; the passenger car in the foreground covers independently moving adult participant 2 from about 2.5 s to 3.1 s; the passenger car in the foreground covers independently moving adult participant 2 from about 4.5 s to 5.0 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 1.7 s to 1.9 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 4.5 s to 5.0 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 3.2 s to 3.4 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 2.3 s to 3.2 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 3.2 s to 3.6 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 2.9 s to 4.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.5 s to 4.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.5 s to 3.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-017-city-combat/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-017-city-combat/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..95964c0ff16b7aca138c9fc85b0352fb2fdd3ce4
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-017-city-combat/follow/mixed/prompt.txt
@@ -0,0 +1,64 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 13> (independently moving adult participant 1), <Subject 14> (independently moving adult participant 2), <Subject 15> (independently moving adult participant 3), <Subject 16> (the training opponent moving laterally and returning fire), <Subject 17> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 11> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 10> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 12> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered surface, positioned in the middle of the street. A training rifle with a matte black finish, held by the player. A vintage sedan with a deep navy body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 10> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 11> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 12> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 13> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 14> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 15> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 16> is the training opponent moving laterally and returning fire; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 17> is the second training opponent returning fire from the opposite side; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 17>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 11> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 12> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 17> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered surface, positioned in the middle of the street. A training rifle with a matte black finish, held by the player. A vintage sedan with a deep navy body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the short-lived shot tracer, realized with the materials above. <Subject 11> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 12> is the short-lived shot tracer, realized with the materials above. <Subject 13> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 15> (independently moving adult participant 3) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the training opponent moving laterally and returning fire) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 17> (the second training opponent returning fire from the opposite side) is a skater with shaggy chestnut hair, wearing a charcoal graphic tee, olive baggy jeans, and scuffed black skate shoes, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about one and a half metres along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 11> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and comes to rest about 12 metres from the protagonist. <Subject 11> is one continuous object throughout and comes to rest about 14 metres from the protagonist. The short-lived shot tracer in the foreground covers independently moving adult participant 1 from about 2.3 s to 3.1 s; the protagonist in the foreground covers independently moving adult participant 2 from about 1.7 s to 1.9 s; the passenger car in the foreground covers independently moving adult participant 2 from about 2.5 s to 3.1 s; the passenger car in the foreground covers independently moving adult participant 2 from about 4.5 s to 5.0 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 1.7 s to 1.9 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 4.5 s to 5.0 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 3.2 s to 3.4 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 2.3 s to 3.2 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 3.2 s to 3.6 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 2.9 s to 4.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.5 s to 4.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.5 s to 3.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-017-city-combat/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-017-city-combat/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..02846a6270311451ef5bb8432abf487824cd717c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-017-city-combat/follow/white/prompt.txt
@@ -0,0 +1,64 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 13> (independently moving adult participant 1), <Subject 14> (independently moving adult participant 2), <Subject 15> (independently moving adult participant 3), <Subject 16> (the training opponent moving laterally and returning fire), <Subject 17> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 11> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 10> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 12> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered surface, positioned in the middle of the street. A training rifle with a matte black finish, held by the player. A vintage sedan with a deep navy body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 10> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 11> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 12> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 13> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 14> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 15> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 16> is the training opponent moving laterally and returning fire; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 17> is the second training opponent returning fire from the opposite side; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 17>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 11> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 12> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 17> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered surface, positioned in the middle of the street. A training rifle with a matte black finish, held by the player. A vintage sedan with a deep navy body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the short-lived shot tracer, realized with the materials above. <Subject 11> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 12> is the short-lived shot tracer, realized with the materials above. <Subject 13> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 15> (independently moving adult participant 3) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the training opponent moving laterally and returning fire) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 17> (the second training opponent returning fire from the opposite side) is a skater with shaggy chestnut hair, wearing a charcoal graphic tee, olive baggy jeans, and scuffed black skate shoes, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about one and a half metres along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 11> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and comes to rest about 12 metres from the protagonist. <Subject 11> is one continuous object throughout and comes to rest about 14 metres from the protagonist. The short-lived shot tracer in the foreground covers independently moving adult participant 1 from about 2.3 s to 3.1 s; the protagonist in the foreground covers independently moving adult participant 2 from about 1.7 s to 1.9 s; the passenger car in the foreground covers independently moving adult participant 2 from about 2.5 s to 3.1 s; the passenger car in the foreground covers independently moving adult participant 2 from about 4.5 s to 5.0 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 1.7 s to 1.9 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 4.5 s to 5.0 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 3.2 s to 3.4 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 2.3 s to 3.2 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 3.2 s to 3.6 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 2.9 s to 4.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.5 s to 4.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.5 s to 3.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-018-city-combat/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-018-city-combat/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..734b6bd88db930d28cb4b8c6c9ad3cd281596ceb
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-018-city-combat/follow/colored-obb/prompt.txt
@@ -0,0 +1,62 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 human proxies standing for <Subject 1> (the protagonist), <Subject 12> (independently moving adult participant 1), <Subject 13> (independently moving adult participant 2), <Subject 14> (independently moving adult participant 3), <Subject 15> (the training opponent moving laterally and returning fire), <Subject 16> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 11> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height barrier made of dark, matte composite material with a rough, industrial finish. A training rifle with a matte black receiver, a long barrel, and a detachable magazine, held by the player. A small hover-car with a sleek, aerodynamic body in a metallic silver shade, hovering slightly above the ground with glowing undercarriage lights - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 12> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 13> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 14> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 15> is the training opponent moving laterally and returning fire; in <Video 1> the human proxy tagged #f73677; use the target appearance described below.
+<Subject 16> is the second training opponent returning fire from the opposite side; in <Video 1> the human proxy tagged #36aff7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 16>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height barrier made of dark, matte composite material with a rough, industrial finish. A training rifle with a matte black receiver, a long barrel, and a detachable magazine, held by the player. A small hover-car with a sleek, aerodynamic body in a metallic silver shade, hovering slightly above the ground with glowing undercarriage lights - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult with short black hair, wearing a safety-amber exoskeleton frame over a charcoal jumpsuit with leg braces and heavy boots, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (independently moving adult participant 1) is an adult with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (independently moving adult participant 2) is an adult with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 3) is an adult with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (the training opponent moving laterally and returning fire) is an android with short synthetic silver hair, featuring a gunmetal chassis with amber-lit seams and metal feet, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the second training opponent returning fire from the opposite side) is an adult hidden under a clear helmet, wearing an ivory pressure suit with navy trim and grey suit boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves a short distance along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the vehicle moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; the vehicle moves about half a metre along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; the vehicle moves a short distance along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and ends the clip about 12 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 1.0 s to 1.9 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 4.1 s to 4.7 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 0.9 s to 2.4 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 4.1 s to 5.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 1.2 s to 2.4 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-018-city-combat/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-018-city-combat/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7fb86bb6b1bcde7c3c3f051a5558ddfc8b7feb31
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-018-city-combat/follow/mixed/prompt.txt
@@ -0,0 +1,62 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 12> (independently moving adult participant 1), <Subject 13> (independently moving adult participant 2), <Subject 14> (independently moving adult participant 3), <Subject 15> (the training opponent moving laterally and returning fire), <Subject 16> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 11> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height barrier made of dark, matte composite material with a rough, industrial finish. A training rifle with a matte black receiver, a long barrel, and a detachable magazine, held by the player. A small hover-car with a sleek, aerodynamic body in a metallic silver shade, hovering slightly above the ground with glowing undercarriage lights - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 12> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 13> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 14> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 15> is the training opponent moving laterally and returning fire; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 16> is the second training opponent returning fire from the opposite side; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 16>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height barrier made of dark, matte composite material with a rough, industrial finish. A training rifle with a matte black receiver, a long barrel, and a detachable magazine, held by the player. A small hover-car with a sleek, aerodynamic body in a metallic silver shade, hovering slightly above the ground with glowing undercarriage lights - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult with short black hair, wearing a safety-amber exoskeleton frame over a charcoal jumpsuit with leg braces and heavy boots, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (independently moving adult participant 1) is an adult with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (independently moving adult participant 2) is an adult with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 3) is an adult with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (the training opponent moving laterally and returning fire) is an android with short synthetic silver hair, featuring a gunmetal chassis with amber-lit seams and metal feet, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the second training opponent returning fire from the opposite side) is an adult hidden under a clear helmet, wearing an ivory pressure suit with navy trim and grey suit boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves a short distance along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the vehicle moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; the vehicle moves about half a metre along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; the vehicle moves a short distance along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and ends the clip about 12 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 1.0 s to 1.9 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 4.1 s to 4.7 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 0.9 s to 2.4 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 4.1 s to 5.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 1.2 s to 2.4 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-018-city-combat/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-018-city-combat/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bff2af8106749696b2373683ac2e6d387ce9fb5c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-018-city-combat/follow/white/prompt.txt
@@ -0,0 +1,62 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 12> (independently moving adult participant 1), <Subject 13> (independently moving adult participant 2), <Subject 14> (independently moving adult participant 3), <Subject 15> (the training opponent moving laterally and returning fire), <Subject 16> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 11> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height barrier made of dark, matte composite material with a rough, industrial finish. A training rifle with a matte black receiver, a long barrel, and a detachable magazine, held by the player. A small hover-car with a sleek, aerodynamic body in a metallic silver shade, hovering slightly above the ground with glowing undercarriage lights - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 12> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 13> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 14> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 15> is the training opponent moving laterally and returning fire; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 16> is the second training opponent returning fire from the opposite side; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 16>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height barrier made of dark, matte composite material with a rough, industrial finish. A training rifle with a matte black receiver, a long barrel, and a detachable magazine, held by the player. A small hover-car with a sleek, aerodynamic body in a metallic silver shade, hovering slightly above the ground with glowing undercarriage lights - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult with short black hair, wearing a safety-amber exoskeleton frame over a charcoal jumpsuit with leg braces and heavy boots, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (independently moving adult participant 1) is an adult with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (independently moving adult participant 2) is an adult with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 3) is an adult with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (the training opponent moving laterally and returning fire) is an android with short synthetic silver hair, featuring a gunmetal chassis with amber-lit seams and metal feet, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the second training opponent returning fire from the opposite side) is an adult hidden under a clear helmet, wearing an ivory pressure suit with navy trim and grey suit boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves a short distance along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the vehicle moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; the vehicle moves about half a metre along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; the vehicle moves a short distance along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and ends the clip about 12 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 1.0 s to 1.9 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 4.1 s to 4.7 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 0.9 s to 2.4 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 4.1 s to 5.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 1.2 s to 2.4 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-022-indoor-place/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-022-indoor-place/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b2e502a27d57c40c83abc333e48240a8836a4b92
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-022-indoor-place/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal tubing with a wooden platform, featuring a utilitarian industrial design - at the start of the clip its front toward the camera. A rectangular cargo crate made of weathered cardboard with a matte brown finish, supported by the pickup stand. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal tubing with a wooden platform, featuring a utilitarian industrial design - at the start of the clip its front toward the camera. A rectangular cargo crate made of weathered cardboard with a matte brown finish, supported by the pickup stand. <Subject 1>, the protagonist, has the following target appearance: A realistic adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.8 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 2.6 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.8 s to 1.3 s; the room-lintel in the foreground covers independently moving adult participant 2 from about 2.2 s to 2.5 s; the room-lintel in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 3.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-022-indoor-place/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-022-indoor-place/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..296d5dcb15b3306ee764a14a13e4d7d5641d616e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-022-indoor-place/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal tubing with a wooden platform, featuring a utilitarian industrial design - at the start of the clip its front toward the camera. A rectangular cargo crate made of weathered cardboard with a matte brown finish, supported by the pickup stand. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal tubing with a wooden platform, featuring a utilitarian industrial design - at the start of the clip its front toward the camera. A rectangular cargo crate made of weathered cardboard with a matte brown finish, supported by the pickup stand. <Subject 1>, the protagonist, has the following target appearance: A realistic adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.8 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 2.6 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.8 s to 1.3 s; the room-lintel in the foreground covers independently moving adult participant 2 from about 2.2 s to 2.5 s; the room-lintel in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 3.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-022-indoor-place/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-022-indoor-place/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..016d79e2f72770879f7cb0cb0cc9a2a5c0d0bad2
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-022-indoor-place/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal tubing with a wooden platform, featuring a utilitarian industrial design - at the start of the clip its front toward the camera. A rectangular cargo crate made of weathered cardboard with a matte brown finish, supported by the pickup stand. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal tubing with a wooden platform, featuring a utilitarian industrial design - at the start of the clip its front toward the camera. A rectangular cargo crate made of weathered cardboard with a matte brown finish, supported by the pickup stand. <Subject 1>, the protagonist, has the following target appearance: A realistic adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.8 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 2.6 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.8 s to 1.3 s; the room-lintel in the foreground covers independently moving adult participant 2 from about 2.2 s to 2.5 s; the room-lintel in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 3.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b635f111abc1a45fdaa43a7567918426208bcf06
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, being pushed along the floor, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the object proxy tagged #f79e36; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, being pushed along the floor, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: An adult site engineer with short wavy dark hair, wearing a cream polo, slate trousers, and brown boots, seen only from behind, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is an adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 1.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.5 s to 1.3 s; the protagonist in the foreground covers independently moving adult participant 3 from about 0.8 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.0 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..efea5a178e0e6fad379f6573401060dc29e10061
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, being pushed along the floor, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, being pushed along the floor, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: An adult site engineer with short wavy dark hair, wearing a cream polo, slate trousers, and brown boots, seen only from behind, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is an adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 1.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.5 s to 1.3 s; the protagonist in the foreground covers independently moving adult participant 3 from about 0.8 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.0 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0eb259dd486bc96e94163bf0a9a73af78b066027
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-025-indoor-walk/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, being pushed along the floor, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, being pushed along the floor, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: An adult site engineer with short wavy dark hair, wearing a cream polo, slate trousers, and brown boots, seen only from behind, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is an adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, painted in the same fauvist bold-colour oil manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 1.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.5 s to 1.3 s; the protagonist in the foreground covers independently moving adult participant 3 from about 0.8 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.0 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, fauvist oil look: flat bold areas of unnatural saturated colour, simplified shapes with loose dark contours, joyful clashing hues instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, daylight through tall windows with soft shadows, crisp air-conditioned air, a softly blended distant background; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..39eecf6d135d013a819f7e90af991dc4a993d9d1
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metallic frame and utilitarian design, positioned in the upper centre of the room - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the object proxy tagged #36f796; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36aff7 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the human proxy tagged #e7f736 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #ce36f7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metallic frame and utilitarian design, positioned in the upper centre of the room - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing in the middle left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle far right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, standing in the upper centre - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. The wheeled handcart is one continuous object throughout and ends the clip about 2 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 2.0 s to 2.2 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.5 s to 2.7 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.9 s to 3.2 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.4 s to 3.9 s; the individual furnished desk in the foreground covers independently moving adult participant 2 from about 1.4 s to 3.2 s; the platform-rail-1 in the foreground covers independently moving adult participant 2 from about 1.7 s to 3.9 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f27c0aa56184b5d73582e5018f87bf7d5441adba
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metallic frame and utilitarian design, positioned in the upper centre of the room - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metallic frame and utilitarian design, positioned in the upper centre of the room - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing in the middle left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle far right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, standing in the upper centre - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. The wheeled handcart is one continuous object throughout and ends the clip about 2 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 2.0 s to 2.2 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.5 s to 2.7 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.9 s to 3.2 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.4 s to 3.9 s; the individual furnished desk in the foreground covers independently moving adult participant 2 from about 1.4 s to 3.2 s; the platform-rail-1 in the foreground covers independently moving adult participant 2 from about 1.7 s to 3.9 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4d9d6c124b3f73a163ba29f9cf8c6ce10b8e0b7d
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-026-indoor-stairs/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metallic frame and utilitarian design, positioned in the upper centre of the room - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metallic frame and utilitarian design, positioned in the upper centre of the room - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing in the middle left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle far right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, standing in the upper centre - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. The wheeled handcart is one continuous object throughout and ends the clip about 2 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 2.0 s to 2.2 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.5 s to 2.7 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.9 s to 3.2 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.4 s to 3.9 s; the individual furnished desk in the foreground covers independently moving adult participant 2 from about 1.4 s to 3.2 s; the platform-rail-1 in the foreground covers independently moving adult participant 2 from about 1.7 s to 3.9 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-034-indoor-push/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-034-indoor-push/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..272fad89781759f28e420f468d616b3c1ccaac8a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-034-indoor-push/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned in the middle distance, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front toward the camera. A sturdy wooden cargo crate with a weathered finish and visible grain, sitting on the floor, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> pushing a crate forward with both hands and releasing it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned in the middle distance, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front toward the camera. A sturdy wooden cargo crate with a weathered finish and visible grain, sitting on the floor, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands). <Subject 1>, the protagonist, has the following target appearance: A squat, wide adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing in the foreground, seen only from behind, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a squat, wide adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the background, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a squat, wide adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the background, seen only from behind, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a squat, wide adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the background, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist steps up to the crate and places both hands on it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.967 to 00:04.067, the protagonist pushes the crate forward; the protagonist moves about 5 metres along its route; the crate moves about 5 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.067 to 00:05.000, the protagonist lets go of the crate and it stays where it stopped; the protagonist rises to full height (head height 1.2 to 1.6 m), lowers the hands to the sides, straightens the legs; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist pushes the parcel or cargo crate from 00:00.967 to 00:04.067. The parcel or cargo crate is one continuous object throughout and comes to rest about a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 1.1 s to 1.3 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.8 s to 2.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 2.9 s to 3.1 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-034-indoor-push/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-034-indoor-push/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..822f6f3cbbe4051bc9666e7fdae07484a0c27e35
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-034-indoor-push/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned in the middle distance, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front toward the camera. A sturdy wooden cargo crate with a weathered finish and visible grain, sitting on the floor, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> pushing a crate forward with both hands and releasing it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned in the middle distance, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front toward the camera. A sturdy wooden cargo crate with a weathered finish and visible grain, sitting on the floor, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands). <Subject 1>, the protagonist, has the following target appearance: A squat, wide adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing in the foreground, seen only from behind, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a squat, wide adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the background, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a squat, wide adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the background, seen only from behind, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a squat, wide adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the background, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist steps up to the crate and places both hands on it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.967 to 00:04.067, the protagonist pushes the crate forward; the protagonist moves about 5 metres along its route; the crate moves about 5 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.067 to 00:05.000, the protagonist lets go of the crate and it stays where it stopped; the protagonist rises to full height (head height 1.2 to 1.6 m), lowers the hands to the sides, straightens the legs; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist pushes the parcel or cargo crate from 00:00.967 to 00:04.067. The parcel or cargo crate is one continuous object throughout and comes to rest about a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 1.1 s to 1.3 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.8 s to 2.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 2.9 s to 3.1 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-034-indoor-push/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-034-indoor-push/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6ffab96614807575860ad97fa790bae1ef5dea3e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-034-indoor-push/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned in the middle distance, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front toward the camera. A sturdy wooden cargo crate with a weathered finish and visible grain, sitting on the floor, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> pushing a crate forward with both hands and releasing it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned in the middle distance, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front toward the camera. A sturdy wooden cargo crate with a weathered finish and visible grain, sitting on the floor, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands). <Subject 1>, the protagonist, has the following target appearance: A squat, wide adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing in the foreground, seen only from behind, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a squat, wide adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the background, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a squat, wide adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the background, seen only from behind, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a squat, wide adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the background, drawn in the same loose watercolour cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist steps up to the crate and places both hands on it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.967 to 00:04.067, the protagonist pushes the crate forward; the protagonist moves about 5 metres along its route; the crate moves about 5 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.067 to 00:05.000, the protagonist lets go of the crate and it stays where it stopped; the protagonist rises to full height (head height 1.2 to 1.6 m), lowers the hands to the sides, straightens the legs; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist pushes the parcel or cargo crate from 00:00.967 to 00:04.067. The parcel or cargo crate is one continuous object throughout and comes to rest about a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 1.1 s to 1.3 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.8 s to 2.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 2.9 s to 3.1 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, loose watercolour cartoon look: soft bleeding colour washes inside light ink outlines, white paper showing through, playful simplified figures, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, dim lamp-lit interior with a bright doorway, a slight interior haze, painted storybook scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..dad81d65ee8975637f9fc7412e0919752a55bd34
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a deep navy body, featuring tinted windows, a front windshield, and heavy rubber tires, positioned in the distance - at the start of the clip its front away from the camera. A small open-top buggy with a rust-red body, featuring a roll cage, a steering wheel, and four off-road tires, positioned in the foreground - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the object proxy tagged #36aff7 that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #f73677 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking up the entry steps into a stationary passenger bus, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a deep navy body, featuring tinted windows, a front windshield, and heavy rubber tires, positioned in the distance - at the start of the clip its front away from the camera. A small open-top buggy with a rust-red body, featuring a roll cage, a steering wheel, and four off-road tires, positioned in the foreground - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A compact, stocky student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a compact, stocky security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a compact, stocky tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a compact, stocky bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.667, the protagonist walks up the entry steps of the stationary bus; the protagonist moves about 2 metres along its route, rises a short distance; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the vehicle moves about 2 metres along its route. From 00:01.667 to 00:05.000, the protagonist moves along the aisle inside the bus; the protagonist moves about 4 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 5 metres along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:01.667 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 4 metres from the protagonist. The enclosed passenger bus in the foreground covers the protagonist from about 1.2 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 2 from about 0.7 s to 1.4 s; the track--8-1 in the foreground covers independently moving adult participant 2 from about 4.7 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 2 from about 4.3 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 3 from about 4.1 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0a2eb725ef3d0c1d4af1583a38603399b68d6b90
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a deep navy body, featuring tinted windows, a front windshield, and heavy rubber tires, positioned in the distance - at the start of the clip its front away from the camera. A small open-top buggy with a rust-red body, featuring a roll cage, a steering wheel, and four off-road tires, positioned in the foreground - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking up the entry steps into a stationary passenger bus, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a deep navy body, featuring tinted windows, a front windshield, and heavy rubber tires, positioned in the distance - at the start of the clip its front away from the camera. A small open-top buggy with a rust-red body, featuring a roll cage, a steering wheel, and four off-road tires, positioned in the foreground - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A compact, stocky student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a compact, stocky security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a compact, stocky tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a compact, stocky bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.667, the protagonist walks up the entry steps of the stationary bus; the protagonist moves about 2 metres along its route, rises a short distance; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the vehicle moves about 2 metres along its route. From 00:01.667 to 00:05.000, the protagonist moves along the aisle inside the bus; the protagonist moves about 4 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 5 metres along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:01.667 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 4 metres from the protagonist. The enclosed passenger bus in the foreground covers the protagonist from about 1.2 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 2 from about 0.7 s to 1.4 s; the track--8-1 in the foreground covers independently moving adult participant 2 from about 4.7 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 2 from about 4.3 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 3 from about 4.1 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5bce6a13a142fb8d85e6d1689c4dd56f6118d154
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-035-indoor-busBoard/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a deep navy body, featuring tinted windows, a front windshield, and heavy rubber tires, positioned in the distance - at the start of the clip its front away from the camera. A small open-top buggy with a rust-red body, featuring a roll cage, a steering wheel, and four off-road tires, positioned in the foreground - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking up the entry steps into a stationary passenger bus, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a deep navy body, featuring tinted windows, a front windshield, and heavy rubber tires, positioned in the distance - at the start of the clip its front away from the camera. A small open-top buggy with a rust-red body, featuring a roll cage, a steering wheel, and four off-road tires, positioned in the foreground - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A compact, stocky student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a compact, stocky security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a compact, stocky tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a compact, stocky bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.667, the protagonist walks up the entry steps of the stationary bus; the protagonist moves about 2 metres along its route, rises a short distance; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the vehicle moves about 2 metres along its route. From 00:01.667 to 00:05.000, the protagonist moves along the aisle inside the bus; the protagonist moves about 4 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 5 metres along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:01.667 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 4 metres from the protagonist. The enclosed passenger bus in the foreground covers the protagonist from about 1.2 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 2 from about 0.7 s to 1.4 s; the track--8-1 in the foreground covers independently moving adult participant 2 from about 4.7 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 2 from about 4.3 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 3 from about 4.1 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..01990f7ee9a6a53f4c134e0bcec682bac9fccacc
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/colored-obb/prompt.txt
@@ -0,0 +1,32 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist); nothing else in <Video 1> is a person or an interaction object. The proxy tagged #36aff7 (person) never enters the frame of this shot: no figure or object appears for it and none must be added.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: An adult field explorer with long braided dark hair, wearing a field shirt with rolled sleeves, cargo trousers, and leather boots, standing upright in the center of the hall, seen only from behind, drawn in the same graphite pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ebbd8ac0550a55ec660e464b7968804e3090bfaa
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/mixed/prompt.txt
@@ -0,0 +1,32 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: An adult field explorer with long braided dark hair, wearing a field shirt with rolled sleeves, cargo trousers, and leather boots, standing upright in the center of the hall, seen only from behind, drawn in the same graphite pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..69863cbee5aea8999c7bf2d2f451735a1d25a9a9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-037-indoor-stairsDown/follow/white/prompt.txt
@@ -0,0 +1,32 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: An adult field explorer with long braided dark hair, wearing a field shirt with rolled sleeves, cargo trousers, and leather boots, standing upright in the center of the hall, seen only from behind, drawn in the same graphite pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, faint dust motes in the light beams, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..15747d6931f1848e210dd690757c580e3ac44835
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the movable chair), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and wooden slats, standing empty in the middle distance to the left, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front toward the camera. A standard wooden chair with a simple backrest and four legs, rendered with cross-hatched shading to indicate its solid form, positioned directly behind the protagonist, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the movable chair; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #f79e36 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> lifting a chair, carrying it forward and setting it down, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and wooden slats, standing empty in the middle distance to the left, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front toward the camera. A standard wooden chair with a simple backrest and four legs, rendered with cross-hatched shading to indicate its solid form, positioned directly behind the protagonist, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines). <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing centrally in the foreground, seen only from behind, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the movable chair, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing to the left of the frame, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly hair, wearing a floral shirt, shorts, and sandals, standing in the middle distance to the right, seen only from behind, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a mechanic with hair tied back, wearing work overalls over a cream tee and steel-toe boots, standing further back on the left side, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the chair and takes hold of it; the protagonist raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the chair off the ground; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the chair forward; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the chair to the ground; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the chair rests on the ground and the protagonist lets go of it; the protagonist lowers the hands; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. Recorded relations: the protagonist holds the movable chair from 00:00.733 to 00:04.567. The movable chair is one continuous object throughout and comes to rest about a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 3 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 2.8 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 4.6 s to 5.0 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 2.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.2 s to 3.2 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b2b71aba83d3c448134c808510c68d2626f3f4ca
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the movable chair), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and wooden slats, standing empty in the middle distance to the left, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front toward the camera. A standard wooden chair with a simple backrest and four legs, rendered with cross-hatched shading to indicate its solid form, positioned directly behind the protagonist, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the movable chair; in <Video 1> the black solid chair proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> lifting a chair, carrying it forward and setting it down, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and wooden slats, standing empty in the middle distance to the left, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front toward the camera. A standard wooden chair with a simple backrest and four legs, rendered with cross-hatched shading to indicate its solid form, positioned directly behind the protagonist, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines). <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing centrally in the foreground, seen only from behind, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the movable chair, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing to the left of the frame, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly hair, wearing a floral shirt, shorts, and sandals, standing in the middle distance to the right, seen only from behind, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a mechanic with hair tied back, wearing work overalls over a cream tee and steel-toe boots, standing further back on the left side, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the chair and takes hold of it; the protagonist raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the chair off the ground; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the chair forward; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the chair to the ground; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the chair rests on the ground and the protagonist lets go of it; the protagonist lowers the hands; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. Recorded relations: the protagonist holds the movable chair from 00:00.733 to 00:04.567. The movable chair is one continuous object throughout and comes to rest about a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 3 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 2.8 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 4.6 s to 5.0 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 2.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.2 s to 3.2 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..04b7b1b6f680ee84ad830287d7afb4a80546f56f
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-038-indoor-carryChair/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the movable chair), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and wooden slats, standing empty in the middle distance to the left, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front toward the camera. A standard wooden chair with a simple backrest and four legs, rendered with cross-hatched shading to indicate its solid form, positioned directly behind the protagonist, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the movable chair; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> lifting a chair, carrying it forward and setting it down, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and wooden slats, standing empty in the middle distance to the left, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front toward the camera. A standard wooden chair with a simple backrest and four legs, rendered with cross-hatched shading to indicate its solid form, positioned directly behind the protagonist, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines). <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing centrally in the foreground, seen only from behind, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the movable chair, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing to the left of the frame, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly hair, wearing a floral shirt, shorts, and sandals, standing in the middle distance to the right, seen only from behind, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a mechanic with hair tied back, wearing work overalls over a cream tee and steel-toe boots, standing further back on the left side, drawn in the same graphite architectural sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the chair and takes hold of it; the protagonist raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the chair off the ground; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the chair forward; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the chair to the ground; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the chair rests on the ground and the protagonist lets go of it; the protagonist lowers the hands; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. Recorded relations: the protagonist holds the movable chair from 00:00.733 to 00:04.567. The movable chair is one continuous object throughout and comes to rest about a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 3 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 2.8 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 4.6 s to 5.0 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 2.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.2 s to 3.2 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite architectural sketch look: precise ruled edges on buildings, light construction lines left visible, controlled parallel hatching, cool grey graphite on smooth paper, drawn over the full 3D world so every recorded shape stays readable; a cool grey tonal palette, mixed window daylight and warm practical lamps, crisp air-conditioned air, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..532851f6481465d88460e028822c2835bea94805
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/colored-obb/prompt.txt
@@ -0,0 +1,54 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 1> (the protagonist), <Subject 12> (the training opponent moving laterally and returning fire), <Subject 13> (the second training opponent returning fire from the opposite side) and 5 object proxies standing for <Subject 3> (the solid waist-height cover used in both exchanges), <Subject 4> (the solid cover that stops incoming shots), <Subject 5> (the training rifle with stock, receiver, barrel and magazine), <Subject 8> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 9> is not in frame at 00:00.000 and enters at about 0.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height cover made of stacked crates, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). A training rifle with a stock, receiver, barrel, and magazine, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the solid waist-height cover used in both exchanges; in <Video 1> the solid proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid cover that stops incoming shots; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 7> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 8> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 9> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 12> is the training opponent moving laterally and returning fire; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 13> is the second training opponent returning fire from the opposite side; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 13>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 9> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height cover made of stacked crates, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). A training rifle with a stock, receiver, barrel, and magazine, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in a ready stance, seen only from behind, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 4> is the solid cover that stops incoming shots, realized with the materials above. <Subject 5> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 6> is the short-lived shot tracer, realized with the materials above. <Subject 7> is the short-lived muzzle flash, realized with the materials above. <Subject 8> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 9> is the short-lived shot tracer, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (the training opponent moving laterally and returning fire) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in a ready stance, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (the second training opponent returning fire from the opposite side) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, and steel-toe boots, standing in a ready stance, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; the training opponent moving laterally and returning fire moves about half a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), lowers the hands, straightens the legs; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 5> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 8> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. <Subject 5> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 8> is one continuous object throughout and comes to rest about 6 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fed1052a50962ae7b4495a26aa705de7bd5f3ef6
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/mixed/prompt.txt
@@ -0,0 +1,54 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 12> (the training opponent moving laterally and returning fire), <Subject 13> (the second training opponent returning fire from the opposite side) and 5 object proxies standing for <Subject 3> (the solid waist-height cover used in both exchanges), <Subject 4> (the solid cover that stops incoming shots), <Subject 5> (the training rifle with stock, receiver, barrel and magazine), <Subject 8> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 9> is not in frame at 00:00.000 and enters at about 0.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height cover made of stacked crates, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). A training rifle with a stock, receiver, barrel, and magazine, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the solid waist-height cover used in both exchanges; in <Video 1> the black solid obstacle proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid cover that stops incoming shots; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 5> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 6> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 7> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 8> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 9> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 12> is the training opponent moving laterally and returning fire; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 13> is the second training opponent returning fire from the opposite side; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 13>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 9> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height cover made of stacked crates, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). A training rifle with a stock, receiver, barrel, and magazine, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in a ready stance, seen only from behind, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 4> is the solid cover that stops incoming shots, realized with the materials above. <Subject 5> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 6> is the short-lived shot tracer, realized with the materials above. <Subject 7> is the short-lived muzzle flash, realized with the materials above. <Subject 8> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 9> is the short-lived shot tracer, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (the training opponent moving laterally and returning fire) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in a ready stance, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (the second training opponent returning fire from the opposite side) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, and steel-toe boots, standing in a ready stance, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; the training opponent moving laterally and returning fire moves about half a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), lowers the hands, straightens the legs; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 5> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 8> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. <Subject 5> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 8> is one continuous object throughout and comes to rest about 6 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..46d6081e6f23dee72c6c5a879b9efb37d60df3ac
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-039-indoor-combat/follow/white/prompt.txt
@@ -0,0 +1,54 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 12> (the training opponent moving laterally and returning fire), <Subject 13> (the second training opponent returning fire from the opposite side) and 5 object proxies standing for <Subject 3> (the solid waist-height cover used in both exchanges), <Subject 4> (the solid cover that stops incoming shots), <Subject 5> (the training rifle with stock, receiver, barrel and magazine), <Subject 8> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 9> is not in frame at 00:00.000 and enters at about 0.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height cover made of stacked crates, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). A training rifle with a stock, receiver, barrel, and magazine, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the solid waist-height cover used in both exchanges; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid cover that stops incoming shots; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 7> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 8> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 9> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 12> is the training opponent moving laterally and returning fire; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 13> is the second training opponent returning fire from the opposite side; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 13>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 9> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height cover made of stacked crates, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). A training rifle with a stock, receiver, barrel, and magazine, rendered in cool desaturated tones with a matte finish, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing). <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in a ready stance, seen only from behind, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 4> is the solid cover that stops incoming shots, realized with the materials above. <Subject 5> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 6> is the short-lived shot tracer, realized with the materials above. <Subject 7> is the short-lived muzzle flash, realized with the materials above. <Subject 8> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 9> is the short-lived shot tracer, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (the training opponent moving laterally and returning fire) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in a ready stance, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (the second training opponent returning fire from the opposite side) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, and steel-toe boots, standing in a ready stance, drawn as an anime character in the same painterly feature-film anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; the training opponent moving laterally and returning fire moves about half a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), lowers the hands, straightens the legs; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 5> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 8> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. <Subject 5> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 8> is one continuous object throughout and comes to rest about 6 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — painterly feature-film anime look: lush fully painted backgrounds with detailed light and clouds, cel-shaded characters with fine linework, rich atmospheric depth; a rich saturated jewel-tone palette, warm tungsten interior light with pools of shadow, faint dust motes in the light beams; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-041-nature-jump/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-041-nature-jump/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2f21526334ca0745b3db10ad7b72126b667ad214
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-041-nature-jump/follow/colored-obb/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the fallen log across the path); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, weathered timber lying horizontally across the path, its surface dark and slick with moisture, covered in patches of pale moss and lichen, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the fallen log across the path; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> approaching a fallen log, jumping over it and landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, weathered timber lying horizontally across the path, its surface dark and slick with moisture, covered in patches of pale moss and lichen, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors). <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing an olive ranger shirt, tan trousers, and brown boots, standing in the foreground with a steady posture, seen only from behind, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the fallen log across the path, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing slightly to the left of the path, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back on the right side, seen only from behind, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the distance near the center, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist moves toward the fallen log; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:01.267 to 00:02.433, the protagonist leaps over the log; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.433 to 00:05.000, the protagonist lands beyond the log and continues forward; the protagonist moves about 2 metres along its route, is back on the ground; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. The tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 2.8 s to 3.6 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 3.2 s to 3.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-041-nature-jump/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-041-nature-jump/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..926e15a77b01f3a555aca880a2591bac1d6bb4dc
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-041-nature-jump/follow/mixed/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the fallen log across the path); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, weathered timber lying horizontally across the path, its surface dark and slick with moisture, covered in patches of pale moss and lichen, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the fallen log across the path; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> approaching a fallen log, jumping over it and landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, weathered timber lying horizontally across the path, its surface dark and slick with moisture, covered in patches of pale moss and lichen, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors). <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing an olive ranger shirt, tan trousers, and brown boots, standing in the foreground with a steady posture, seen only from behind, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the fallen log across the path, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing slightly to the left of the path, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back on the right side, seen only from behind, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the distance near the center, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist moves toward the fallen log; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:01.267 to 00:02.433, the protagonist leaps over the log; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.433 to 00:05.000, the protagonist lands beyond the log and continues forward; the protagonist moves about 2 metres along its route, is back on the ground; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. The tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 2.8 s to 3.6 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 3.2 s to 3.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-041-nature-jump/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-041-nature-jump/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..33362f3390adffa5045c6bf0c42ee9e76dfbdf27
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-041-nature-jump/follow/white/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the fallen log across the path); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, weathered timber lying horizontally across the path, its surface dark and slick with moisture, covered in patches of pale moss and lichen, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the fallen log across the path; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> approaching a fallen log, jumping over it and landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, weathered timber lying horizontally across the path, its surface dark and slick with moisture, covered in patches of pale moss and lichen, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors). <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing an olive ranger shirt, tan trousers, and brown boots, standing in the foreground with a steady posture, seen only from behind, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the fallen log across the path, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing slightly to the left of the path, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back on the right side, seen only from behind, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is a field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the distance near the center, drawn in the same sanguine conte-crayon sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist moves toward the fallen log; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:01.267 to 00:02.433, the protagonist leaps over the log; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.433 to 00:05.000, the protagonist lands beyond the log and continues forward; the protagonist moves about 2 metres along its route, is back on the ground; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. The tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 2.8 s to 3.6 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 3.2 s to 3.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, sanguine conte sketch look: warm red-brown chalk strokes, smudged mid-tones, white chalk highlights on toned tan paper, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, blue-hour dusk with a few warm practical lights, thin fog hugging the ground, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4a02a875a7402b756fe4172e7add91859b9a219f
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person receiving the object), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered, matte finish and visible grain, held by the protagonist, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered, matte finish and visible grain, held by the protagonist, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds). <Subject 1>, the protagonist, has the following target appearance: An adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the middle ground, seen only from behind, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing near the crate, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is an adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing on the far left, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back in the distance, seen only from behind, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is not in frame at 00:00.000 and enters the frame at about 0.3 s; from then on it is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, currently outside the frame, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d2751b1cc486980b49f29d2e5e4861d2140e6f17
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person receiving the object), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered, matte finish and visible grain, held by the protagonist, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered, matte finish and visible grain, held by the protagonist, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds). <Subject 1>, the protagonist, has the following target appearance: An adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the middle ground, seen only from behind, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing near the crate, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is an adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing on the far left, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back in the distance, seen only from behind, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is not in frame at 00:00.000 and enters the frame at about 0.3 s; from then on it is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, currently outside the frame, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6c2e6d072f00f6c5c8b7116fcb2bf0603ac537b3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-042-nature-handoff/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person receiving the object), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered, matte finish and visible grain, held by the protagonist, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered, matte finish and visible grain, held by the protagonist, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds). <Subject 1>, the protagonist, has the following target appearance: An adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the middle ground, seen only from behind, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing near the crate, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is an adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing on the far left, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back in the distance, seen only from behind, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is not in frame at 00:00.000 and enters the frame at about 0.3 s; from then on it is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, currently outside the frame, painted in the same plein-air oil sketch manner (character design: figures with warm glazed skin and bold cloth folds), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-043-nature-walk/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-043-nature-walk/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5a48c404e1653587241395b1740dd24cef4de3cc
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-043-nature-walk/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2), <Subject 5> (independently moving adult participant 3); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the foreground with a focused posture, seen only from behind, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> (independently moving adult participant 1) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing slightly to the left and further back, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the middle distance to the right, seen only from behind, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing furthest back on the left, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 2.4 s to 3.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-043-nature-walk/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-043-nature-walk/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d143661c17f3e3f6b6d92b2ba61ef994a0455152
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-043-nature-walk/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2), <Subject 5> (independently moving adult participant 3); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the foreground with a focused posture, seen only from behind, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> (independently moving adult participant 1) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing slightly to the left and further back, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the middle distance to the right, seen only from behind, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing furthest back on the left, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 2.4 s to 3.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-043-nature-walk/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-043-nature-walk/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8365940651aabe2dc5844cfb50ebdc3eadb5fbd
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-043-nature-walk/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2), <Subject 5> (independently moving adult participant 3); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the foreground with a focused posture, seen only from behind, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> (independently moving adult participant 1) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing slightly to the left and further back, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the middle distance to the right, seen only from behind, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing furthest back on the left, painted in the same blue-green mineral-pigment ink wash manner (character design: figures as a few confident brush gestures) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 2.4 s to 3.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, blue-green mineral ink look: ink contours filled with azure and malachite mineral washes, layered distant planes, ornamental rocks and clouds, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a moody deep-shadow palette with luminous highlights, blue-hour dusk with a few warm practical lights, fine rain and wet reflective ground, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a3f9610e0e0c497110546a04c1536662c3d84640
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/colored-obb/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2), <Subject 5> (independently moving adult participant 3); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking along a trail while looking around and glancing back, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> (independently moving adult participant 1) is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing to the middle left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the upper right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 3) is an adult fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the upper centre distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.633, the protagonist stands on the trail and takes in the surroundings. From 00:00.633 to 00:03.967, the protagonist walks along the trail while looking around at the surroundings; the protagonist moves about 7 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route. From 00:03.967 to 00:04.767, the protagonist turns back to face forward along the trail; the protagonist moves about half a metre along its route; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly along the trail. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8de60fd70b1d93cd5206c1b35b253151698e2cf5
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/mixed/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2), <Subject 5> (independently moving adult participant 3); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking along a trail while looking around and glancing back, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> (independently moving adult participant 1) is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing to the middle left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the upper right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 3) is an adult fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the upper centre distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.633, the protagonist stands on the trail and takes in the surroundings. From 00:00.633 to 00:03.967, the protagonist walks along the trail while looking around at the surroundings; the protagonist moves about 7 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route. From 00:03.967 to 00:04.767, the protagonist turns back to face forward along the trail; the protagonist moves about half a metre along its route; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly along the trail. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e5f21143ecca25e452478dcdb078e17debd5f463
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-044-nature-camera/first-person/white/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2), <Subject 5> (independently moving adult participant 3); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking along a trail while looking around and glancing back, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> (independently moving adult participant 1) is an adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing to the middle left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the upper right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 3) is an adult fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the upper centre distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.633, the protagonist stands on the trail and takes in the surroundings. From 00:00.633 to 00:03.967, the protagonist walks along the trail while looking around at the surroundings; the protagonist moves about 7 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route. From 00:03.967 to 00:04.767, the protagonist turns back to face forward along the trail; the protagonist moves about half a metre along its route; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly along the trail. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-048-nature-ride/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-048-nature-ride/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b43043ea6db65e8cf5a506ae0e852c58d5df43fa
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-048-nature-ride/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the ridden saddle horse carrying the protagonist), <Subject 4> (riderless herd horse 1), <Subject 5> (riderless herd horse 2); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Animals in this shot: riderless herd horse 1 (left of the frame): a chestnut horse with a reddish-brown coat, no saddle and no rider, grazing in the middle distance, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. Riderless herd horse 2 (right of the frame): a bay horse with a dark brown coat, no saddle and no rider, standing alert in the middle right of the frame, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. The ridden horse (foreground centre): a bay riding horse with a rich dark brown coat, a leather saddle and tack, carrying the protagonist in the saddle and walking forward through the pasture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the saddle horse that the protagonist rides for the whole clip; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is riderless herd horse 1; in <Video 1> the object proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is riderless herd horse 2; in <Video 1> the object proxy tagged #3ef736 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse past other horses, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Animals in this shot: riderless herd horse 1 (left of the frame): a chestnut horse with a reddish-brown coat, no saddle and no rider, grazing in the middle distance, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. Riderless herd horse 2 (right of the frame): a bay horse with a dark brown coat, no saddle and no rider, standing alert in the middle right of the frame, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. The ridden horse (foreground centre): a bay riding horse with a rich dark brown coat, a leather saddle and tack, carrying the protagonist in the saddle and walking forward through the pasture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, seated upright in the saddle of the ridden horse with the reins in hand, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the ridden horse, realized with the materials above. <Subject 4> is riderless herd horse 1, realized with the materials above. <Subject 5> is riderless herd horse 2, realized with the materials above. <Subject 6> (independently moving adult participant 1) is a farmer with grey hair under a flat cap, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, observing the scene with interest, seen only from behind, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing casually in the background, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward past the other horses; the protagonist moves about 4 metres along its route; the ridden horse moves about 4 metres along its route with the protagonist in the saddle; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; herd horse 1 moves about one and a half metres along its route; herd horse 2 moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward past the other horses; the protagonist moves about 4 metres along its route; the ridden horse moves about 4 metres along its route with the protagonist in the saddle; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; herd horse 1 moves about one and a half metres along its route; herd horse 2 moves about one and a half metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. The ridden horse stays under the protagonist for the whole clip. Herd horse 1 is one continuous animal throughout and ends the clip about 7 metres from the protagonist. Herd horse 2 is one continuous animal throughout and ends the clip about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 1.3 s to 2.0 s; the ridden horse in the foreground covers independently moving adult participant 3 from about 1.2 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-048-nature-ride/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-048-nature-ride/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bbd143c2bf64b7b418807c33e2f836bcb7fce2d8
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-048-nature-ride/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the ridden saddle horse carrying the protagonist), <Subject 4> (riderless herd horse 1), <Subject 5> (riderless herd horse 2); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Animals in this shot: riderless herd horse 1 (left of the frame): a chestnut horse with a reddish-brown coat, no saddle and no rider, grazing in the middle distance, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. Riderless herd horse 2 (right of the frame): a bay horse with a dark brown coat, no saddle and no rider, standing alert in the middle right of the frame, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. The ridden horse (foreground centre): a bay riding horse with a rich dark brown coat, a leather saddle and tack, carrying the protagonist in the saddle and walking forward through the pasture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the saddle horse that the protagonist rides for the whole clip; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is riderless herd horse 1; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is riderless herd horse 2; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse past other horses, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Animals in this shot: riderless herd horse 1 (left of the frame): a chestnut horse with a reddish-brown coat, no saddle and no rider, grazing in the middle distance, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. Riderless herd horse 2 (right of the frame): a bay horse with a dark brown coat, no saddle and no rider, standing alert in the middle right of the frame, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. The ridden horse (foreground centre): a bay riding horse with a rich dark brown coat, a leather saddle and tack, carrying the protagonist in the saddle and walking forward through the pasture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, seated upright in the saddle of the ridden horse with the reins in hand, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the ridden horse, realized with the materials above. <Subject 4> is riderless herd horse 1, realized with the materials above. <Subject 5> is riderless herd horse 2, realized with the materials above. <Subject 6> (independently moving adult participant 1) is a farmer with grey hair under a flat cap, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, observing the scene with interest, seen only from behind, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing casually in the background, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward past the other horses; the protagonist moves about 4 metres along its route; the ridden horse moves about 4 metres along its route with the protagonist in the saddle; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; herd horse 1 moves about one and a half metres along its route; herd horse 2 moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward past the other horses; the protagonist moves about 4 metres along its route; the ridden horse moves about 4 metres along its route with the protagonist in the saddle; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; herd horse 1 moves about one and a half metres along its route; herd horse 2 moves about one and a half metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. The ridden horse stays under the protagonist for the whole clip. Herd horse 1 is one continuous animal throughout and ends the clip about 7 metres from the protagonist. Herd horse 2 is one continuous animal throughout and ends the clip about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 1.3 s to 2.0 s; the ridden horse in the foreground covers independently moving adult participant 3 from about 1.2 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-048-nature-ride/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-048-nature-ride/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..da0c476ad444a1481198e49823bf42344c8095d3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-048-nature-ride/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the ridden saddle horse carrying the protagonist), <Subject 4> (riderless herd horse 1), <Subject 5> (riderless herd horse 2); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Animals in this shot: riderless herd horse 1 (left of the frame): a chestnut horse with a reddish-brown coat, no saddle and no rider, grazing in the middle distance, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. Riderless herd horse 2 (right of the frame): a bay horse with a dark brown coat, no saddle and no rider, standing alert in the middle right of the frame, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. The ridden horse (foreground centre): a bay riding horse with a rich dark brown coat, a leather saddle and tack, carrying the protagonist in the saddle and walking forward through the pasture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the saddle horse that the protagonist rides for the whole clip; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is riderless herd horse 1; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is riderless herd horse 2; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse past other horses, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Animals in this shot: riderless herd horse 1 (left of the frame): a chestnut horse with a reddish-brown coat, no saddle and no rider, grazing in the middle distance, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. Riderless herd horse 2 (right of the frame): a bay horse with a dark brown coat, no saddle and no rider, standing alert in the middle right of the frame, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. The ridden horse (foreground centre): a bay riding horse with a rich dark brown coat, a leather saddle and tack, carrying the protagonist in the saddle and walking forward through the pasture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, seated upright in the saddle of the ridden horse with the reins in hand, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the ridden horse, realized with the materials above. <Subject 4> is riderless herd horse 1, realized with the materials above. <Subject 5> is riderless herd horse 2, realized with the materials above. <Subject 6> (independently moving adult participant 1) is a farmer with grey hair under a flat cap, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, observing the scene with interest, seen only from behind, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing casually in the background, painted in the same post-impressionist swirling oil manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward past the other horses; the protagonist moves about 4 metres along its route; the ridden horse moves about 4 metres along its route with the protagonist in the saddle; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; herd horse 1 moves about one and a half metres along its route; herd horse 2 moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward past the other horses; the protagonist moves about 4 metres along its route; the ridden horse moves about 4 metres along its route with the protagonist in the saddle; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; herd horse 1 moves about one and a half metres along its route; herd horse 2 moves about one and a half metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. The ridden horse stays under the protagonist for the whole clip. Herd horse 1 is one continuous animal throughout and ends the clip about 7 metres from the protagonist. Herd horse 2 is one continuous animal throughout and ends the clip about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 1.3 s to 2.0 s; the ridden horse in the foreground covers independently moving adult participant 3 from about 1.2 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, post-impressionist oil look: thick rhythmic swirling strokes, exaggerated vibrant complementary colours, outlines suggested by dark paint ridges, instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, gentle wind moving loose fabric and foliage, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-050-nature-herd/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-050-nature-herd/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e5cdf9c65f4bf902ae2aa0a2862689c9e66a0915
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-050-nature-herd/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the saddle horse), <Subject 4> (the saddle horse), <Subject 5> (the saddle horse); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A chestnut horse with a reddish-brown coat and a simple saddle, standing in the lower far left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A grey draught horse with a dappled grey coat, standing in the middle right, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A bay riding horse with a dark brown coat and a leather saddle, standing in the lower centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the saddle horse; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the saddle horse; in <Video 1> the object proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the saddle horse; in <Video 1> the object proxy tagged #3ef736 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse while horses and pedestrians move nearby, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A chestnut horse with a reddish-brown coat and a simple saddle, standing in the lower far left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A grey draught horse with a dappled grey coat, standing in the middle right, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A bay riding horse with a dark brown coat and a leather saddle, standing in the lower centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A rugged adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing in the middle centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the saddle horse, realized with the materials above. <Subject 4> is the saddle horse, realized with the materials above. <Subject 5> is the saddle horse, realized with the materials above. <Subject 6> (independently moving adult participant 1) is a rugged adult farmer with grey hair under a flat cap, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing in the middle left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a rugged adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the middle centre, seen only from behind, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a rugged adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the middle centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward while the other horses and the pedestrians move nearby; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward among the moving horses and pedestrians; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about half a metre along its route; the horse moves about half a metre along its route; the horse moves about 2 metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 7 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. The paddock fence in the foreground covers independently moving adult participant 1 from about 3.8 s to 4.0 s; the barn outside the paddock in the foreground covers independently moving adult participant 1 from about 3.5 s to 3.8 s; the saddle horse in the foreground covers independently moving adult participant 1 from about 2.6 s to 3.0 s; the saddle horse in the foreground covers independently moving adult participant 1 from about 3.7 s to 4.8 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.9 s to 1.2 s; the protagonist in the foreground covers independently moving adult participant 2 from about 3.9 s to 4.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-050-nature-herd/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-050-nature-herd/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..96c8d6ca397799f68b1d457c133903458ddf5987
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-050-nature-herd/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the saddle horse), <Subject 4> (the saddle horse), <Subject 5> (the saddle horse); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A chestnut horse with a reddish-brown coat and a simple saddle, standing in the lower far left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A grey draught horse with a dappled grey coat, standing in the middle right, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A bay riding horse with a dark brown coat and a leather saddle, standing in the lower centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the saddle horse; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the saddle horse; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the saddle horse; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse while horses and pedestrians move nearby, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A chestnut horse with a reddish-brown coat and a simple saddle, standing in the lower far left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A grey draught horse with a dappled grey coat, standing in the middle right, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A bay riding horse with a dark brown coat and a leather saddle, standing in the lower centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A rugged adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing in the middle centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the saddle horse, realized with the materials above. <Subject 4> is the saddle horse, realized with the materials above. <Subject 5> is the saddle horse, realized with the materials above. <Subject 6> (independently moving adult participant 1) is a rugged adult farmer with grey hair under a flat cap, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing in the middle left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a rugged adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the middle centre, seen only from behind, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a rugged adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the middle centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward while the other horses and the pedestrians move nearby; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward among the moving horses and pedestrians; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about half a metre along its route; the horse moves about half a metre along its route; the horse moves about 2 metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 7 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. The paddock fence in the foreground covers independently moving adult participant 1 from about 3.8 s to 4.0 s; the barn outside the paddock in the foreground covers independently moving adult participant 1 from about 3.5 s to 3.8 s; the saddle horse in the foreground covers independently moving adult participant 1 from about 2.6 s to 3.0 s; the saddle horse in the foreground covers independently moving adult participant 1 from about 3.7 s to 4.8 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.9 s to 1.2 s; the protagonist in the foreground covers independently moving adult participant 2 from about 3.9 s to 4.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-050-nature-herd/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-050-nature-herd/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d9a2f688936f89a81a1814600533d70a7d74d3eb
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-050-nature-herd/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the saddle horse), <Subject 4> (the saddle horse), <Subject 5> (the saddle horse); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A chestnut horse with a reddish-brown coat and a simple saddle, standing in the lower far left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A grey draught horse with a dappled grey coat, standing in the middle right, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A bay riding horse with a dark brown coat and a leather saddle, standing in the lower centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the saddle horse; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the saddle horse; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the saddle horse; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse while horses and pedestrians move nearby, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A chestnut horse with a reddish-brown coat and a simple saddle, standing in the lower far left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A grey draught horse with a dappled grey coat, standing in the middle right, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A bay riding horse with a dark brown coat and a leather saddle, standing in the lower centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A rugged adult park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing in the middle centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the saddle horse, realized with the materials above. <Subject 4> is the saddle horse, realized with the materials above. <Subject 5> is the saddle horse, realized with the materials above. <Subject 6> (independently moving adult participant 1) is a rugged adult farmer with grey hair under a flat cap, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing in the middle left, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a rugged adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the middle centre, seen only from behind, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a rugged adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the middle centre, painted in the same soft ink on silk manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward while the other horses and the pedestrians move nearby; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward among the moving horses and pedestrians; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about half a metre along its route; the horse moves about half a metre along its route; the horse moves about 2 metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 7 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. The paddock fence in the foreground covers independently moving adult participant 1 from about 3.8 s to 4.0 s; the barn outside the paddock in the foreground covers independently moving adult participant 1 from about 3.5 s to 3.8 s; the saddle horse in the foreground covers independently moving adult participant 1 from about 2.6 s to 3.0 s; the saddle horse in the foreground covers independently moving adult participant 1 from about 3.7 s to 4.8 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.9 s to 1.2 s; the protagonist in the foreground covers independently moving adult participant 2 from about 3.9 s to 4.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, ink-on-silk look: softly diffused ink edges soaked into a warm silk weave, muted tea-coloured ground, restrained tonal shading, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a low-saturation earthy palette, misty morning backlight, clear dry air, short dry scratchy strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..aaec6e119f8c4e6e675f03d98ab25ed9132ed0e9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 human proxies standing for <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The lift is a sturdy metal platform with a matte finish, positioned at the lower centre of the frame, its edges defined by quick ink lines and subtle watercolour shading, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). The switch is a small control panel with a brushed metal surface, located slightly to the right of the protagonist, its details hinted at with fine ink strokes, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the human proxy tagged #6636f7; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the human proxy tagged #3ef736; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is activating a lift platform, ascending with it and stepping onto the upper landing (the camera itself rises with the platform; <Subject 1> is never visible as a figure), involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The lift is a sturdy metal platform with a matte finish, positioned at the lower centre of the frame, its edges defined by quick ink lines and subtle watercolour shading, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). The switch is a small control panel with a brushed metal surface, located slightly to the right of the protagonist, its details hinted at with fine ink strokes, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). <Subject 1>, the protagonist, has the following target appearance: the camera wearer, never shown as a figure: this is a first-person view from the protagonist's own eyes, so the protagonist's body, head and face are never visible; at most the protagonist's own hand may enter the lower edge of the frame when pressing the lift control, drawn in the same fine-liner ink pen sketch manner. <Subject 3> (independently moving adult participant 1) is an adult hiker with short dark hair, dressed in a shell jacket, charcoal trousers, and hiking boots, stays small in the middle distance, slightly to the right, well beyond the platform, seen only from behind, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is an adult student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, stands small and far back, near the upper centre of the frame, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist presses the control on the lift platform; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. From 00:00.967 to 00:03.767, the lift platform rises with the protagonist standing on it, so the camera rises about one and a half metres; the protagonist rises about one and a half metres; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route. From 00:03.767 to 00:05.000, the protagonist steps off the platform onto the upper landing; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. The upper-platform in the foreground covers independently moving adult participant 1 from about 0.2 s to 1.7 s; the upper-platform in the foreground covers independently moving adult participant 2 from about 0.0 s to 3.0 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e011983ca32e0af01bb79456804d4a9b92f9554f
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 blue mannequin proxies standing for <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The lift is a sturdy metal platform with a matte finish, positioned at the lower centre of the frame, its edges defined by quick ink lines and subtle watercolour shading, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). The switch is a small control panel with a brushed metal surface, located slightly to the right of the protagonist, its details hinted at with fine ink strokes, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is activating a lift platform, ascending with it and stepping onto the upper landing (the camera itself rises with the platform; <Subject 1> is never visible as a figure), involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The lift is a sturdy metal platform with a matte finish, positioned at the lower centre of the frame, its edges defined by quick ink lines and subtle watercolour shading, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). The switch is a small control panel with a brushed metal surface, located slightly to the right of the protagonist, its details hinted at with fine ink strokes, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). <Subject 1>, the protagonist, has the following target appearance: the camera wearer, never shown as a figure: this is a first-person view from the protagonist's own eyes, so the protagonist's body, head and face are never visible; at most the protagonist's own hand may enter the lower edge of the frame when pressing the lift control, drawn in the same fine-liner ink pen sketch manner. <Subject 3> (independently moving adult participant 1) is an adult hiker with short dark hair, dressed in a shell jacket, charcoal trousers, and hiking boots, stays small in the middle distance, slightly to the right, well beyond the platform, seen only from behind, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is an adult student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, stands small and far back, near the upper centre of the frame, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist presses the control on the lift platform; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. From 00:00.967 to 00:03.767, the lift platform rises with the protagonist standing on it, so the camera rises about one and a half metres; the protagonist rises about one and a half metres; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route. From 00:03.767 to 00:05.000, the protagonist steps off the platform onto the upper landing; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. The upper-platform in the foreground covers independently moving adult participant 1 from about 0.2 s to 1.7 s; the upper-platform in the foreground covers independently moving adult participant 2 from about 0.0 s to 3.0 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9f97bad3ab5286d3283abaa251722c876782b4f0
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-052-nature-lift/first-person/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 grey block-figure proxies standing for <Subject 3> (independently moving adult participant 1), <Subject 4> (independently moving adult participant 2); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The lift is a sturdy metal platform with a matte finish, positioned at the lower centre of the frame, its edges defined by quick ink lines and subtle watercolour shading, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). The switch is a small control panel with a brushed metal surface, located slightly to the right of the protagonist, its details hinted at with fine ink strokes, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 4> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is activating a lift platform, ascending with it and stepping onto the upper landing (the camera itself rises with the platform; <Subject 1> is never visible as a figure), involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The lift is a sturdy metal platform with a matte finish, positioned at the lower centre of the frame, its edges defined by quick ink lines and subtle watercolour shading, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). The switch is a small control panel with a brushed metal surface, located slightly to the right of the protagonist, its details hinted at with fine ink strokes, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces). <Subject 1>, the protagonist, has the following target appearance: the camera wearer, never shown as a figure: this is a first-person view from the protagonist's own eyes, so the protagonist's body, head and face are never visible; at most the protagonist's own hand may enter the lower edge of the frame when pressing the lift control, drawn in the same fine-liner ink pen sketch manner. <Subject 3> (independently moving adult participant 1) is an adult hiker with short dark hair, dressed in a shell jacket, charcoal trousers, and hiking boots, stays small in the middle distance, slightly to the right, well beyond the platform, seen only from behind, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 4> (independently moving adult participant 2) is an adult student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, stands small and far back, near the upper centre of the frame, drawn in the same fine-liner ink pen sketch manner (character design: figures carefully shaded with clear faces) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist presses the control on the lift platform; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. From 00:00.967 to 00:03.767, the lift platform rises with the protagonist standing on it, so the camera rises about one and a half metres; the protagonist rises about one and a half metres; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route. From 00:03.767 to 00:05.000, the protagonist steps off the platform onto the upper landing; the protagonist moves about 2 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route. The upper-platform in the foreground covers independently moving adult participant 1 from about 0.2 s to 1.7 s; the upper-platform in the foreground covers independently moving adult participant 2 from about 0.0 s to 3.0 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, fine-liner ink pen sketch look: thin black ink contour lines, dense cross-hatching for shadow, white paper, no colour except a single wash tint, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, thin fog hugging the ground, background rendered with the same care as the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7be0bb2b7b0b7c750182d86d79efc8f9296486b4
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/colored-obb/prompt.txt
@@ -0,0 +1,50 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 9> (independently moving diver 1), <Subject 10> (independently moving diver 2), <Subject 11> (independently moving diver 3) and 5 animal proxies (fish) standing for <Subject 4> (the reef fish following the leading fish in a loose school), <Subject 5> (the reef fish following the leading fish in a loose school), <Subject 6> (the reef fish following the leading fish in a loose school), <Subject 7> (the reef fish following the leading fish in a loose school), <Subject 8> (the reef fish following the leading fish in a loose school) and 1 object proxy standing for <Subject 3> (the hinged door); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 5> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.7 s; <Subject 7> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 8> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 9> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 2.7 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rusted metal door frame, part of a shipwreck, stands in the middle centre of the frame, its surface textured with marine growth and encrustations, appearing weathered and aged. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper centre of the frame, appearing - at the start of the clip its front away from the camera. Another small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the top right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the hinged door; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 5> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the object proxy tagged #3ef736; use the target appearance described below.
+<Subject 6> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.7 s; in <Video 1> the object proxy tagged #f73677; use the target appearance described below.
+<Subject 7> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the object proxy tagged #36aff7; use the target appearance described below.
+<Subject 8> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the object proxy tagged #e7f736; use the target appearance described below.
+<Subject 9> is independently moving diver 1: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 10> is independently moving diver 2; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 11> is independently moving diver 3: not in frame at 00:00.000; it enters the frame at about 2.7 s; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> opening a submerged hatch while swimming, involving <Subject 3> to <Subject 11>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (enters [Shot 1] at about 2.7 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 11> (enters [Shot 1] at about 2.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rusted metal door frame, part of a shipwreck, stands in the middle centre of the frame, its surface textured with marine growth and encrustations, appearing weathered and aged. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper centre of the frame, appearing - at the start of the clip its front away from the camera. Another small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the top right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult diver with dark hair, wearing a charcoal wetsuit with slate panels and fins, stands in the middle centre of the frame, appearing sharp and detailed against the blurred background, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the hinged door, realized with the materials above. <Subject 4> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 5> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 6> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 7> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 8> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 9> (independently moving diver 1) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, is positioned in the upper centre of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving diver 2) is a realistic adult fisher, wearing a rust waterproof jacket and olive waders, is positioned in the upper far right of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 11> (independently moving diver 3) is not in frame at 00:00.000 and enters the frame at about 2.7 s; from then on it is a realistic adult mechanic with auburn hair tied back, wearing navy work overalls, is positioned in the upper right of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist swims up to the submerged hatch; the protagonist moves about half a metre along its route, raises the hands to chest height; independently moving diver 1 moves about half a metre along its route; independently moving diver 2 moves about half a metre along its route; independently moving diver 3 moves about half a metre along its route; the fish moves about a metre along its route; the fish moves about half a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about half a metre along its route. From 00:00.967 to 00:02.167, the protagonist works the hatch open while treading water; the protagonist lowers the hands to the sides; independently moving diver 1 moves about half a metre along its route; independently moving diver 2 moves about a metre along its route; independently moving diver 3 moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route. From 00:02.167 to 00:05.000, the hatch stands open and the protagonist swims on; the protagonist moves about 2 metres along its route, lowers the body by about half a metre (head height 2.9 to 2.5 m); independently moving diver 1 moves about one and a half metres along its route; independently moving diver 2 moves about 2 metres along its route; independently moving diver 3 moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route. Recorded relations: the protagonist opens the hinged door from 00:00.000 to 00:02.167; <Subject 5>, <Subject 6>, <Subject 7> and <Subject 8> follow <Subject 4> from 00:00.000 to 00:05.000; the protagonist opened the hinged door from 00:02.167 to 00:05.000. <Subject 4> is one continuous object throughout and ends the clip about 7 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 11 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 8> is one continuous object throughout and ends the clip about 16 metres from the protagonist. The protagonist in the foreground covers independently moving diver 1 from about 2.7 s to 2.9 s; the protagonist in the foreground covers independently moving diver 1 from about 4.0 s to 4.2 s; the wreck-wall--3.2 in the foreground covers independently moving diver 1 from about 0.0 s to 2.3 s; the wreck-front-1 in the foreground covers independently moving diver 1 from about 0.0 s to 2.5 s; the hinged door in the foreground covers independently moving diver 1 from about 0.0 s to 0.4 s; the wreck-wall-3.2 in the foreground covers independently moving diver 2 from about 2.5 s to 5.0 s; the wreck-front-1 in the foreground covers independently moving diver 2 from about 2.5 s to 5.0 s; the wreck-wall-3.2 in the foreground covers independently moving diver 3 from about 0.9 s to 2.2 s; the wreck-front-1 in the foreground covers independently moving diver 3 from about 0.0 s to 2.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..08d46481d6a1ab88da13f00b38e43027e0d34ddb
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/mixed/prompt.txt
@@ -0,0 +1,50 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 9> (independently moving diver 1), <Subject 10> (independently moving diver 2), <Subject 11> (independently moving diver 3) and 5 animal proxies (fish) standing for <Subject 4> (the reef fish following the leading fish in a loose school), <Subject 5> (the reef fish following the leading fish in a loose school), <Subject 6> (the reef fish following the leading fish in a loose school), <Subject 7> (the reef fish following the leading fish in a loose school), <Subject 8> (the reef fish following the leading fish in a loose school) and 1 object proxy standing for <Subject 3> (the hinged door); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 5> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.7 s; <Subject 7> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 8> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 9> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 2.7 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rusted metal door frame, part of a shipwreck, stands in the middle centre of the frame, its surface textured with marine growth and encrustations, appearing weathered and aged. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper centre of the frame, appearing - at the start of the clip its front away from the camera. Another small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the top right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the hinged door; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 5> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 6> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.7 s; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 7> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 8> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 9> is independently moving diver 1: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 10> is independently moving diver 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 11> is independently moving diver 3: not in frame at 00:00.000; it enters the frame at about 2.7 s; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> opening a submerged hatch while swimming, involving <Subject 3> to <Subject 11>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (enters [Shot 1] at about 2.7 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 11> (enters [Shot 1] at about 2.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rusted metal door frame, part of a shipwreck, stands in the middle centre of the frame, its surface textured with marine growth and encrustations, appearing weathered and aged. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper centre of the frame, appearing - at the start of the clip its front away from the camera. Another small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the top right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult diver with dark hair, wearing a charcoal wetsuit with slate panels and fins, stands in the middle centre of the frame, appearing sharp and detailed against the blurred background, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the hinged door, realized with the materials above. <Subject 4> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 5> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 6> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 7> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 8> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 9> (independently moving diver 1) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, is positioned in the upper centre of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving diver 2) is a realistic adult fisher, wearing a rust waterproof jacket and olive waders, is positioned in the upper far right of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 11> (independently moving diver 3) is not in frame at 00:00.000 and enters the frame at about 2.7 s; from then on it is a realistic adult mechanic with auburn hair tied back, wearing navy work overalls, is positioned in the upper right of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist swims up to the submerged hatch; the protagonist moves about half a metre along its route, raises the hands to chest height; independently moving diver 1 moves about half a metre along its route; independently moving diver 2 moves about half a metre along its route; independently moving diver 3 moves about half a metre along its route; the fish moves about a metre along its route; the fish moves about half a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about half a metre along its route. From 00:00.967 to 00:02.167, the protagonist works the hatch open while treading water; the protagonist lowers the hands to the sides; independently moving diver 1 moves about half a metre along its route; independently moving diver 2 moves about a metre along its route; independently moving diver 3 moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route. From 00:02.167 to 00:05.000, the hatch stands open and the protagonist swims on; the protagonist moves about 2 metres along its route, lowers the body by about half a metre (head height 2.9 to 2.5 m); independently moving diver 1 moves about one and a half metres along its route; independently moving diver 2 moves about 2 metres along its route; independently moving diver 3 moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route. Recorded relations: the protagonist opens the hinged door from 00:00.000 to 00:02.167; <Subject 5>, <Subject 6>, <Subject 7> and <Subject 8> follow <Subject 4> from 00:00.000 to 00:05.000; the protagonist opened the hinged door from 00:02.167 to 00:05.000. <Subject 4> is one continuous object throughout and ends the clip about 7 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 11 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 8> is one continuous object throughout and ends the clip about 16 metres from the protagonist. The protagonist in the foreground covers independently moving diver 1 from about 2.7 s to 2.9 s; the protagonist in the foreground covers independently moving diver 1 from about 4.0 s to 4.2 s; the wreck-wall--3.2 in the foreground covers independently moving diver 1 from about 0.0 s to 2.3 s; the wreck-front-1 in the foreground covers independently moving diver 1 from about 0.0 s to 2.5 s; the hinged door in the foreground covers independently moving diver 1 from about 0.0 s to 0.4 s; the wreck-wall-3.2 in the foreground covers independently moving diver 2 from about 2.5 s to 5.0 s; the wreck-front-1 in the foreground covers independently moving diver 2 from about 2.5 s to 5.0 s; the wreck-wall-3.2 in the foreground covers independently moving diver 3 from about 0.9 s to 2.2 s; the wreck-front-1 in the foreground covers independently moving diver 3 from about 0.0 s to 2.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..27d43663857a0c116ad8f971c0406c6b36029bd2
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-059-underwater-hatch/follow/white/prompt.txt
@@ -0,0 +1,50 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 9> (independently moving diver 1), <Subject 10> (independently moving diver 2), <Subject 11> (independently moving diver 3) and 5 animal proxies (fish) standing for <Subject 4> (the reef fish following the leading fish in a loose school), <Subject 5> (the reef fish following the leading fish in a loose school), <Subject 6> (the reef fish following the leading fish in a loose school), <Subject 7> (the reef fish following the leading fish in a loose school), <Subject 8> (the reef fish following the leading fish in a loose school) and 1 object proxy standing for <Subject 3> (the hinged door); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 5> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.7 s; <Subject 7> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 8> is not in frame at 00:00.000 and enters at about 2.8 s; <Subject 9> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 2.7 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rusted metal door frame, part of a shipwreck, stands in the middle centre of the frame, its surface textured with marine growth and encrustations, appearing weathered and aged. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper centre of the frame, appearing - at the start of the clip its front away from the camera. Another small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the top right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the hinged door; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.7 s; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 8> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 9> is independently moving diver 1: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 10> is independently moving diver 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 11> is independently moving diver 3: not in frame at 00:00.000; it enters the frame at about 2.7 s; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> opening a submerged hatch while swimming, involving <Subject 3> to <Subject 11>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (enters [Shot 1] at about 2.7 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 11> (enters [Shot 1] at about 2.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rusted metal door frame, part of a shipwreck, stands in the middle centre of the frame, its surface textured with marine growth and encrustations, appearing weathered and aged. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper centre of the frame, appearing - at the start of the clip its front away from the camera. Another small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the top right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, following the leading fish in a loose school, is positioned in the upper right of the frame, appearing - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult diver with dark hair, wearing a charcoal wetsuit with slate panels and fins, stands in the middle centre of the frame, appearing sharp and detailed against the blurred background, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the hinged door, realized with the materials above. <Subject 4> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 5> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 6> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 7> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 8> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 9> (independently moving diver 1) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, is positioned in the upper centre of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving diver 2) is a realistic adult fisher, wearing a rust waterproof jacket and olive waders, is positioned in the upper far right of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 11> (independently moving diver 3) is not in frame at 00:00.000 and enters the frame at about 2.7 s; from then on it is a realistic adult mechanic with auburn hair tied back, wearing navy work overalls, is positioned in the upper right of the frame, appearing, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist swims up to the submerged hatch; the protagonist moves about half a metre along its route, raises the hands to chest height; independently moving diver 1 moves about half a metre along its route; independently moving diver 2 moves about half a metre along its route; independently moving diver 3 moves about half a metre along its route; the fish moves about a metre along its route; the fish moves about half a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about half a metre along its route. From 00:00.967 to 00:02.167, the protagonist works the hatch open while treading water; the protagonist lowers the hands to the sides; independently moving diver 1 moves about half a metre along its route; independently moving diver 2 moves about a metre along its route; independently moving diver 3 moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route; the fish moves about a metre along its route. From 00:02.167 to 00:05.000, the hatch stands open and the protagonist swims on; the protagonist moves about 2 metres along its route, lowers the body by about half a metre (head height 2.9 to 2.5 m); independently moving diver 1 moves about one and a half metres along its route; independently moving diver 2 moves about 2 metres along its route; independently moving diver 3 moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route. Recorded relations: the protagonist opens the hinged door from 00:00.000 to 00:02.167; <Subject 5>, <Subject 6>, <Subject 7> and <Subject 8> follow <Subject 4> from 00:00.000 to 00:05.000; the protagonist opened the hinged door from 00:02.167 to 00:05.000. <Subject 4> is one continuous object throughout and ends the clip about 7 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 11 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 8> is one continuous object throughout and ends the clip about 16 metres from the protagonist. The protagonist in the foreground covers independently moving diver 1 from about 2.7 s to 2.9 s; the protagonist in the foreground covers independently moving diver 1 from about 4.0 s to 4.2 s; the wreck-wall--3.2 in the foreground covers independently moving diver 1 from about 0.0 s to 2.3 s; the wreck-front-1 in the foreground covers independently moving diver 1 from about 0.0 s to 2.5 s; the hinged door in the foreground covers independently moving diver 1 from about 0.0 s to 0.4 s; the wreck-wall-3.2 in the foreground covers independently moving diver 2 from about 2.5 s to 5.0 s; the wreck-front-1 in the foreground covers independently moving diver 2 from about 2.5 s to 5.0 s; the wreck-wall-3.2 in the foreground covers independently moving diver 3 from about 0.9 s to 2.2 s; the wreck-front-1 in the foreground covers independently moving diver 3 from about 0.0 s to 2.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6ad86ff3b00d624a0288c3447972cfed4cc14aca
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/colored-obb/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving diver 1), <Subject 9> (independently moving diver 2), <Subject 10> (independently moving diver 3) and 5 animal proxies (fish) standing for <Subject 3> (the reef fish following the leading fish in a loose school), <Subject 4> (the reef fish following the leading fish in a loose school), <Subject 5> (the reef fish following the leading fish in a loose school), <Subject 6> (the reef fish following the leading fish in a loose school), <Subject 7> (the reef fish following the leading fish in a loose school); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 2.8 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small reef fish with iridescent scales, swimming in a loose school near the center of the frame - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school towards the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school further back in the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school near the wreck's edge - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school in the distant background - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the reef fish following the leading fish in a loose school; in <Video 1> the object proxy tagged #f79e36; use the target appearance described below.
+<Subject 4> is the reef fish following the leading fish in a loose school; in <Video 1> the object proxy tagged #6636f7 that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the reef fish following the leading fish in a loose school; in <Video 1> the object proxy tagged #3ef736; use the target appearance described below.
+<Subject 6> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the object proxy tagged #f73677 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the reef fish following the leading fish in a loose school; in <Video 1> the object proxy tagged #36aff7 that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving diver 1; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 9> is independently moving diver 2; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 10> is independently moving diver 3; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> swimming through an opening into an enclosed shipwreck, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small reef fish with iridescent scales, swimming in a loose school near the center of the frame - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school towards the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school further back in the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school near the wreck's edge - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school in the distant background - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, swimming horizontally through the water, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 4> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 5> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 6> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 7> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 8> (independently moving diver 1) is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, floating with bare feet visible - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving diver 2) is a realistic adult fisher, wearing a rust waterproof jacket and olive waders, appearing to hover near the wreck, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving diver 3) is a realistic adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves and olive cargo trousers, drifting in the distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:03.000, the protagonist swims toward the opening of the shipwreck; the protagonist moves about 6 metres along its route, raises the hands to chest height; independently moving diver 1 moves about one and a half metres along its route; independently moving diver 2 moves about 3 metres along its route; independently moving diver 3 moves about 3 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route. From 00:03.000 to 00:05.000, the protagonist swims on inside the enclosed shipwreck; the protagonist moves about 4 metres along its route, lowers the hands to the sides; independently moving diver 2 moves about 2 metres along its route; independently moving diver 3 moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route. Recorded relations: <Subject 4>, <Subject 5>, <Subject 6> and <Subject 7> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 7 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 10 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 12 metres from the protagonist. The protagonist in the foreground covers independently moving diver 1 from about 2.0 s to 2.5 s; the protagonist in the foreground covers independently moving diver 1 from about 2.6 s to 2.8 s; the wreck-wall--3.2 in the foreground covers independently moving diver 1 from about 0.0 s to 4.0 s; the wreck-wall-3.2 in the foreground covers independently moving diver 2 from about 1.5 s to 1.9 s; the wreck-front-1 in the foreground covers independently moving diver 2 from about 0.0 s to 2.6 s; the wreck-front-1 in the foreground covers independently moving diver 3 from about 0.5 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8df94e61e7956745ed102d2541e99de36ea11154
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/mixed/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving diver 1), <Subject 9> (independently moving diver 2), <Subject 10> (independently moving diver 3) and 5 animal proxies (fish) standing for <Subject 3> (the reef fish following the leading fish in a loose school), <Subject 4> (the reef fish following the leading fish in a loose school), <Subject 5> (the reef fish following the leading fish in a loose school), <Subject 6> (the reef fish following the leading fish in a loose school), <Subject 7> (the reef fish following the leading fish in a loose school); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 2.8 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small reef fish with iridescent scales, swimming in a loose school near the center of the frame - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school towards the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school further back in the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school near the wreck's edge - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school in the distant background - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the reef fish following the leading fish in a loose school; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 4> is the reef fish following the leading fish in a loose school; in <Video 1> the black solid fish proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the reef fish following the leading fish in a loose school; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 6> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the black solid fish proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the reef fish following the leading fish in a loose school; in <Video 1> the black solid fish proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving diver 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 9> is independently moving diver 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 10> is independently moving diver 3; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> swimming through an opening into an enclosed shipwreck, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small reef fish with iridescent scales, swimming in a loose school near the center of the frame - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school towards the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school further back in the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school near the wreck's edge - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school in the distant background - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, swimming horizontally through the water, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 4> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 5> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 6> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 7> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 8> (independently moving diver 1) is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, floating with bare feet visible - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving diver 2) is a realistic adult fisher, wearing a rust waterproof jacket and olive waders, appearing to hover near the wreck, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving diver 3) is a realistic adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves and olive cargo trousers, drifting in the distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:03.000, the protagonist swims toward the opening of the shipwreck; the protagonist moves about 6 metres along its route, raises the hands to chest height; independently moving diver 1 moves about one and a half metres along its route; independently moving diver 2 moves about 3 metres along its route; independently moving diver 3 moves about 3 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route. From 00:03.000 to 00:05.000, the protagonist swims on inside the enclosed shipwreck; the protagonist moves about 4 metres along its route, lowers the hands to the sides; independently moving diver 2 moves about 2 metres along its route; independently moving diver 3 moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route. Recorded relations: <Subject 4>, <Subject 5>, <Subject 6> and <Subject 7> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 7 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 10 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 12 metres from the protagonist. The protagonist in the foreground covers independently moving diver 1 from about 2.0 s to 2.5 s; the protagonist in the foreground covers independently moving diver 1 from about 2.6 s to 2.8 s; the wreck-wall--3.2 in the foreground covers independently moving diver 1 from about 0.0 s to 4.0 s; the wreck-wall-3.2 in the foreground covers independently moving diver 2 from about 1.5 s to 1.9 s; the wreck-front-1 in the foreground covers independently moving diver 2 from about 0.0 s to 2.6 s; the wreck-front-1 in the foreground covers independently moving diver 3 from about 0.5 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..76b014df5b5c4593b31ddae94c0240799a8415aa
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-064-underwater-swimTunnel/follow/white/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving diver 1), <Subject 9> (independently moving diver 2), <Subject 10> (independently moving diver 3) and 5 animal proxies (fish) standing for <Subject 3> (the reef fish following the leading fish in a loose school), <Subject 4> (the reef fish following the leading fish in a loose school), <Subject 5> (the reef fish following the leading fish in a loose school), <Subject 6> (the reef fish following the leading fish in a loose school), <Subject 7> (the reef fish following the leading fish in a loose school); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 2.8 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small reef fish with iridescent scales, swimming in a loose school near the center of the frame - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school towards the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school further back in the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school near the wreck's edge - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school in the distant background - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the reef fish following the leading fish in a loose school; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the reef fish following the leading fish in a loose school; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the reef fish following the leading fish in a loose school; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the reef fish following the leading fish in a loose school: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the reef fish following the leading fish in a loose school; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving diver 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 9> is independently moving diver 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 10> is independently moving diver 3; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> swimming through an opening into an enclosed shipwreck, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 2.8 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small reef fish with iridescent scales, swimming in a loose school near the center of the frame - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school towards the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school further back in the upper right - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school near the wreck's edge - at the start of the clip its front away from the camera. A small reef fish with iridescent scales, swimming in a loose school in the distant background - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, swimming horizontally through the water, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 4> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 5> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 6> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 7> is the reef fish following the leading fish in a loose school, realized with the materials above. <Subject 8> (independently moving diver 1) is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, floating with bare feet visible - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving diver 2) is a realistic adult fisher, wearing a rust waterproof jacket and olive waders, appearing to hover near the wreck, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving diver 3) is a realistic adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves and olive cargo trousers, drifting in the distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:03.000, the protagonist swims toward the opening of the shipwreck; the protagonist moves about 6 metres along its route, raises the hands to chest height; independently moving diver 1 moves about one and a half metres along its route; independently moving diver 2 moves about 3 metres along its route; independently moving diver 3 moves about 3 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route; the fish moves about 2 metres along its route. From 00:03.000 to 00:05.000, the protagonist swims on inside the enclosed shipwreck; the protagonist moves about 4 metres along its route, lowers the hands to the sides; independently moving diver 2 moves about 2 metres along its route; independently moving diver 3 moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route; the fish moves about one and a half metres along its route. Recorded relations: <Subject 4>, <Subject 5>, <Subject 6> and <Subject 7> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 7 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 10 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 12 metres from the protagonist. The protagonist in the foreground covers independently moving diver 1 from about 2.0 s to 2.5 s; the protagonist in the foreground covers independently moving diver 1 from about 2.6 s to 2.8 s; the wreck-wall--3.2 in the foreground covers independently moving diver 1 from about 0.0 s to 4.0 s; the wreck-wall-3.2 in the foreground covers independently moving diver 2 from about 1.5 s to 1.9 s; the wreck-front-1 in the foreground covers independently moving diver 2 from about 0.0 s to 2.6 s; the wreck-front-1 in the foreground covers independently moving diver 3 from about 0.5 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..673943e949a045efa845cd9ca8244642d0556e72
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 1 animal proxy (fish) standing for <Subject 5> (the fish lifted out of the water) and 2 object proxies standing for <Subject 3> (the rod with a reel), <Subject 4> (the fishing line); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A lake shore with a fishing pier, small rowing boats, rocky banks and trees. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small, silvery fish lifted out of the water, its scales glistening with moisture, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip its front away from the camera. A thin, taut monofilament line extending from the rod tip down towards the water surface, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). A slender, dark composite fishing rod with a reel attached, held upright by the protagonist, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the rod with a reel; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the fishing line; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the fish lifted out of the water; in <Video 1> the object proxy tagged #f79e36; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f73677 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36aff7 that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the human proxy tagged #e7f736 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> reeling in a fishing line and lifting a fish out of the water, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A lake shore with a fishing pier, small rowing boats, rocky banks and trees. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small, silvery fish lifted out of the water, its scales glistening with moisture, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip its front away from the camera. A thin, taut monofilament line extending from the rod tip down towards the water surface, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). A slender, dark composite fishing rod with a reel attached, held upright by the protagonist, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). <Subject 1>, the protagonist, has the following target appearance: An adult male fisher, wearing a waterproof jacket, waders, and rubber boots, standing on the pier holding a fishing rod, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the rod with a reel, realized with the materials above. <Subject 4> is the fishing line, realized with the materials above. <Subject 5> is the fish lifted out of the water, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult female tourist with curly hair under a straw hat, wearing a floral shirt, shorts, and sandals, standing on the pier, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an adult male park ranger with short grey hair, wearing a ranger shirt, trousers, and boots, standing further down the pier, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is an adult male hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots, standing in the distance on the pier, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.167, the protagonist reels in the fishing line with the rod; the fish moves about 3 metres along its route, rises about 2 metres; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route. From 00:04.167 to 00:05.000, the protagonist lifts the hooked fish out of the water on the line; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. Recorded relations: the protagonist holds the rod with a reel from 00:00.000 to 00:05.000. The fish lifted out of the water is one continuous object throughout and comes to rest about 2 metres from the protagonist. The continuous water surface in the foreground covers independently moving adult participant 3 from about 1.5 s to 3.7 s; the continuous water surface in the foreground covers independently moving adult participant 3 from about 4.2 s to 4.5 s; the continuous water surface in the foreground covers independently moving adult participant 3 from about 4.7 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9c0b05b1fed302dbbb029eb017a2337be778efc2
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 1 animal proxy (fish) standing for <Subject 5> (the fish lifted out of the water) and 2 object proxies standing for <Subject 3> (the rod with a reel), <Subject 4> (the fishing line); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A lake shore with a fishing pier, small rowing boats, rocky banks and trees. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small, silvery fish lifted out of the water, its scales glistening with moisture, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip its front away from the camera. A thin, taut monofilament line extending from the rod tip down towards the water surface, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). A slender, dark composite fishing rod with a reel attached, held upright by the protagonist, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the rod with a reel; in <Video 1> the black solid fishing-rod proxy shape; use the target appearance described below.
+<Subject 4> is the fishing line; in <Video 1> the black solid fishing-line proxy shape; use the target appearance described below.
+<Subject 5> is the fish lifted out of the water; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> reeling in a fishing line and lifting a fish out of the water, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A lake shore with a fishing pier, small rowing boats, rocky banks and trees. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small, silvery fish lifted out of the water, its scales glistening with moisture, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip its front away from the camera. A thin, taut monofilament line extending from the rod tip down towards the water surface, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). A slender, dark composite fishing rod with a reel attached, held upright by the protagonist, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). <Subject 1>, the protagonist, has the following target appearance: An adult male fisher, wearing a waterproof jacket, waders, and rubber boots, standing on the pier holding a fishing rod, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the rod with a reel, realized with the materials above. <Subject 4> is the fishing line, realized with the materials above. <Subject 5> is the fish lifted out of the water, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult female tourist with curly hair under a straw hat, wearing a floral shirt, shorts, and sandals, standing on the pier, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an adult male park ranger with short grey hair, wearing a ranger shirt, trousers, and boots, standing further down the pier, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is an adult male hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots, standing in the distance on the pier, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.167, the protagonist reels in the fishing line with the rod; the fish moves about 3 metres along its route, rises about 2 metres; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route. From 00:04.167 to 00:05.000, the protagonist lifts the hooked fish out of the water on the line; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. Recorded relations: the protagonist holds the rod with a reel from 00:00.000 to 00:05.000. The fish lifted out of the water is one continuous object throughout and comes to rest about 2 metres from the protagonist. The continuous water surface in the foreground covers independently moving adult participant 3 from about 1.5 s to 3.7 s; the continuous water surface in the foreground covers independently moving adult participant 3 from about 4.2 s to 4.5 s; the continuous water surface in the foreground covers independently moving adult participant 3 from about 4.7 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..41961b5c3f27c0cd3dd9dd3fea73968009d8f206
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-065-underwater-fish/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 1 animal proxy (fish) standing for <Subject 5> (the fish lifted out of the water) and 2 object proxies standing for <Subject 3> (the rod with a reel), <Subject 4> (the fishing line); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A lake shore with a fishing pier, small rowing boats, rocky banks and trees. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small, silvery fish lifted out of the water, its scales glistening with moisture, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip its front away from the camera. A thin, taut monofilament line extending from the rod tip down towards the water surface, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). A slender, dark composite fishing rod with a reel attached, held upright by the protagonist, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the rod with a reel; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the fishing line; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the fish lifted out of the water; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> reeling in a fishing line and lifting a fish out of the water, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A lake shore with a fishing pier, small rowing boats, rocky banks and trees. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small, silvery fish lifted out of the water, its scales glistening with moisture, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip its front away from the camera. A thin, taut monofilament line extending from the rod tip down towards the water surface, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). A slender, dark composite fishing rod with a reel attached, held upright by the protagonist, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies). <Subject 1>, the protagonist, has the following target appearance: An adult male fisher, wearing a waterproof jacket, waders, and rubber boots, standing on the pier holding a fishing rod, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the rod with a reel, realized with the materials above. <Subject 4> is the fishing line, realized with the materials above. <Subject 5> is the fish lifted out of the water, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult female tourist with curly hair under a straw hat, wearing a floral shirt, shorts, and sandals, standing on the pier, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an adult male park ranger with short grey hair, wearing a ranger shirt, trousers, and boots, standing further down the pier, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is an adult male hiker with short dark hair, wearing a shell jacket, charcoal trousers, and hiking boots, standing in the distance on the pier, seen only from behind, painted in the same wet-on-wet misty ink wash manner (character design: figures with fine outlined faces and loose washed bodies) - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.167, the protagonist reels in the fishing line with the rod; the fish moves about 3 metres along its route, rises about 2 metres; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route. From 00:04.167 to 00:05.000, the protagonist lifts the hooked fish out of the water on the line; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. Recorded relations: the protagonist holds the rod with a reel from 00:00.000 to 00:05.000. The fish lifted out of the water is one continuous object throughout and comes to rest about 2 metres from the protagonist. The continuous water surface in the foreground covers independently moving adult participant 3 from about 1.5 s to 3.7 s; the continuous water surface in the foreground covers independently moving adult participant 3 from about 4.2 s to 4.5 s; the continuous water surface in the foreground covers independently moving adult participant 3 from about 4.7 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, wet-on-wet misty ink wash: soft bleeding ink washes, layered grey mist between planes, minimal linework, pale paper tone and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key grey tonal palette, bright shallow-water light, suspended silt particles, a faint paper grain across the whole image; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9d3592cdbd818f6bf1e1ceff971399a638acadb9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 5> (the independently moving adult participant) and 2 object proxies standing for <Subject 3> (the open rowing boat), <Subject 4> (the open rowing boat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: Open lake water seen from the water surface by a swimmer: the camera stays at water level, small rowing boats pass in the distance and a low rocky shoreline with trees lies far away on the horizon; there is no pier, jetty or dock in view. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden skiff with a natural wood finish, moving slowly across the open water on the left, far from the camera - at the start of the clip its front away from the camera. A fishing dory with a matte grey finish, moving slowly across the open water on the right, far from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the open rowing boat; in <Video 1> the object proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is the open rowing boat; in <Video 1> the solid proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the independently moving adult participant; in <Video 1> the human proxy tagged #6636f7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is swimming near passing boats and other people, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open lake water seen from the water surface by a swimmer: the camera stays at water level, small rowing boats pass in the distance and a low rocky shoreline with trees lies far away on the horizon; there is no pier, jetty or dock in view. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden skiff with a natural wood finish, moving slowly across the open water on the left, far from the camera - at the start of the clip its front away from the camera. A fishing dory with a matte grey finish, moving slowly across the open water on the right, far from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the open rowing boat, realized with the materials above. <Subject 4> is the open rowing boat, realized with the materials above. <Subject 5> (the independently moving adult participant) is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, swimming in the water ahead of the camera with only the head and shoulders above the surface, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist swims forward near the passing boats and the other swimmers; the protagonist moves about 4 metres along its route; the independently moving adult participant moves about 2 metres along its route; the boat moves about one and a half metres along its route; the boat moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps swimming forward among the boats and swimmers; the protagonist moves about 4 metres along its route; the independently moving adult participant moves about 2 metres along its route; the boat moves about one and a half metres along its route; the boat moves about one and a half metres along its route. <Subject 3> is one continuous object throughout and ends the clip about 8 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 8 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1ea6fa6c0a4fe0142e6afd7abf83c73ee12acca5
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 5> (the independently moving adult participant) and 2 object proxies standing for <Subject 3> (the open rowing boat), <Subject 4> (the open rowing boat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: Open lake water seen from the water surface by a swimmer: the camera stays at water level, small rowing boats pass in the distance and a low rocky shoreline with trees lies far away on the horizon; there is no pier, jetty or dock in view. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden skiff with a natural wood finish, moving slowly across the open water on the left, far from the camera - at the start of the clip its front away from the camera. A fishing dory with a matte grey finish, moving slowly across the open water on the right, far from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the open rowing boat; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is the open rowing boat; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the independently moving adult participant; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is swimming near passing boats and other people, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open lake water seen from the water surface by a swimmer: the camera stays at water level, small rowing boats pass in the distance and a low rocky shoreline with trees lies far away on the horizon; there is no pier, jetty or dock in view. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden skiff with a natural wood finish, moving slowly across the open water on the left, far from the camera - at the start of the clip its front away from the camera. A fishing dory with a matte grey finish, moving slowly across the open water on the right, far from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the open rowing boat, realized with the materials above. <Subject 4> is the open rowing boat, realized with the materials above. <Subject 5> (the independently moving adult participant) is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, swimming in the water ahead of the camera with only the head and shoulders above the surface, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist swims forward near the passing boats and the other swimmers; the protagonist moves about 4 metres along its route; the independently moving adult participant moves about 2 metres along its route; the boat moves about one and a half metres along its route; the boat moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps swimming forward among the boats and swimmers; the protagonist moves about 4 metres along its route; the independently moving adult participant moves about 2 metres along its route; the boat moves about one and a half metres along its route; the boat moves about one and a half metres along its route. <Subject 3> is one continuous object throughout and ends the clip about 8 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 8 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..36e98848bbccbe74231ba63887eb1963ef291448
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-066-underwater-crowdWater/first-person/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 5> (the independently moving adult participant) and 2 object proxies standing for <Subject 3> (the open rowing boat), <Subject 4> (the open rowing boat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: Open lake water seen from the water surface by a swimmer: the camera stays at water level, small rowing boats pass in the distance and a low rocky shoreline with trees lies far away on the horizon; there is no pier, jetty or dock in view. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden skiff with a natural wood finish, moving slowly across the open water on the left, far from the camera - at the start of the clip its front away from the camera. A fishing dory with a matte grey finish, moving slowly across the open water on the right, far from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the open rowing boat; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is the open rowing boat; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the independently moving adult participant; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is swimming near passing boats and other people, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open lake water seen from the water surface by a swimmer: the camera stays at water level, small rowing boats pass in the distance and a low rocky shoreline with trees lies far away on the horizon; there is no pier, jetty or dock in view. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden skiff with a natural wood finish, moving slowly across the open water on the left, far from the camera - at the start of the clip its front away from the camera. A fishing dory with a matte grey finish, moving slowly across the open water on the right, far from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the open rowing boat, realized with the materials above. <Subject 4> is the open rowing boat, realized with the materials above. <Subject 5> (the independently moving adult participant) is a realistic adult swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, swimming in the water ahead of the camera with only the head and shoulders above the surface, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist swims forward near the passing boats and the other swimmers; the protagonist moves about 4 metres along its route; the independently moving adult participant moves about 2 metres along its route; the boat moves about one and a half metres along its route; the boat moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps swimming forward among the boats and swimmers; the protagonist moves about 4 metres along its route; the independently moving adult participant moves about 2 metres along its route; the boat moves about one and a half metres along its route; the boat moves about one and a half metres along its route. <Subject 3> is one continuous object throughout and ends the clip about 8 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 8 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-067-sky-plane/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-067-sky-plane/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..98599c45e403a02c043a8fcb2816217865ffb7ad
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-067-sky-plane/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object. <Subject 1> is not in frame at 00:00.000 and enters at about 1.4 s.
+<Subject 1> is the protagonist: not in frame at 00:00.000; it enters the frame at about 1.4 s; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished silver fuselage, dark grey wings, and clear cockpit glazing, flying steadily with its propeller spinning - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte charcoal fuselage, dark grey wings, and clear cockpit glazing, maintaining a steady flight path at a distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the object proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (enters [Shot 1] at about 1.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished silver fuselage, dark grey wings, and clear cockpit glazing, flying steadily with its propeller spinning - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte charcoal fuselage, dark grey wings, and clear cockpit glazing, maintaining a steady flight path at a distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing upright with a calm, observant posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.900, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 19 metres along its route. From 00:00.900 to 00:03.167, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 48 metres along its route, drops about a metre; the vehicle moves about 48 metres along its route, drops about a metre; the vehicle moves about 49 metres along its route, drops about a metre. From 00:03.167 to 00:03.567, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 8 metres along its route, drops about half a metre; the vehicle moves about 8 metres along its route, drops about half a metre; the vehicle moves about 8 metres along its route, drops about half a metre. From 00:03.567 to 00:04.667, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 22 metres along its route, drops about 2 metres; the vehicle moves about 22 metres along its route, drops about 2 metres; the vehicle moves about 23 metres along its route, drops about 2 metres. From 00:04.667 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. The piloted propeller aircraft in the foreground covers the protagonist from about 0.0 s to 1.4 s; the piloted propeller aircraft in the foreground covers the protagonist from about 2.9 s to 4.1 s; the piloted propeller aircraft in the foreground covers the protagonist from about 4.5 s to 5.0 s. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-067-sky-plane/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-067-sky-plane/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..febc476dafe1f6ea0450864762b0a591232291cb
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-067-sky-plane/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object. <Subject 1> is not in frame at 00:00.000 and enters at about 1.4 s.
+<Subject 1> is the protagonist: not in frame at 00:00.000; it enters the frame at about 1.4 s; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished silver fuselage, dark grey wings, and clear cockpit glazing, flying steadily with its propeller spinning - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte charcoal fuselage, dark grey wings, and clear cockpit glazing, maintaining a steady flight path at a distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (enters [Shot 1] at about 1.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished silver fuselage, dark grey wings, and clear cockpit glazing, flying steadily with its propeller spinning - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte charcoal fuselage, dark grey wings, and clear cockpit glazing, maintaining a steady flight path at a distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing upright with a calm, observant posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.900, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 19 metres along its route. From 00:00.900 to 00:03.167, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 48 metres along its route, drops about a metre; the vehicle moves about 48 metres along its route, drops about a metre; the vehicle moves about 49 metres along its route, drops about a metre. From 00:03.167 to 00:03.567, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 8 metres along its route, drops about half a metre; the vehicle moves about 8 metres along its route, drops about half a metre; the vehicle moves about 8 metres along its route, drops about half a metre. From 00:03.567 to 00:04.667, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 22 metres along its route, drops about 2 metres; the vehicle moves about 22 metres along its route, drops about 2 metres; the vehicle moves about 23 metres along its route, drops about 2 metres. From 00:04.667 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. The piloted propeller aircraft in the foreground covers the protagonist from about 0.0 s to 1.4 s; the piloted propeller aircraft in the foreground covers the protagonist from about 2.9 s to 4.1 s; the piloted propeller aircraft in the foreground covers the protagonist from about 4.5 s to 5.0 s. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-067-sky-plane/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-067-sky-plane/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bf5c031f61addd689d199d2babcfe5a78ab58b0a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-067-sky-plane/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object. <Subject 1> is not in frame at 00:00.000 and enters at about 1.4 s.
+<Subject 1> is the protagonist: not in frame at 00:00.000; it enters the frame at about 1.4 s; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished silver fuselage, dark grey wings, and clear cockpit glazing, flying steadily with its propeller spinning - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte charcoal fuselage, dark grey wings, and clear cockpit glazing, maintaining a steady flight path at a distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (enters [Shot 1] at about 1.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished silver fuselage, dark grey wings, and clear cockpit glazing, flying steadily with its propeller spinning - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte charcoal fuselage, dark grey wings, and clear cockpit glazing, maintaining a steady flight path at a distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing upright with a calm, observant posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.900, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 19 metres along its route. From 00:00.900 to 00:03.167, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 48 metres along its route, drops about a metre; the vehicle moves about 48 metres along its route, drops about a metre; the vehicle moves about 49 metres along its route, drops about a metre. From 00:03.167 to 00:03.567, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 8 metres along its route, drops about half a metre; the vehicle moves about 8 metres along its route, drops about half a metre; the vehicle moves about 8 metres along its route, drops about half a metre. From 00:03.567 to 00:04.667, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 22 metres along its route, drops about 2 metres; the vehicle moves about 22 metres along its route, drops about 2 metres; the vehicle moves about 23 metres along its route, drops about 2 metres. From 00:04.667 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. The piloted propeller aircraft in the foreground covers the protagonist from about 0.0 s to 1.4 s; the piloted propeller aircraft in the foreground covers the protagonist from about 2.9 s to 4.1 s; the piloted propeller aircraft in the foreground covers the protagonist from about 4.5 s to 5.0 s. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-068-sky-plane/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-068-sky-plane/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a1b1435bd761a976b40e8d1d3282e29531c9b800
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-068-sky-plane/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a matte charcoal fuselage, polished aluminium wings, and clear cockpit glazing, flying through the air, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage, polished aluminium wings, and clear cockpit glazing, maintaining a steady lateral separation from the main aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the object proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a matte charcoal fuselage, polished aluminium wings, and clear cockpit glazing, flying through the air, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage, polished aluminium wings, and clear cockpit glazing, maintaining a steady lateral separation from the main aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing confidently in the cockpit of the aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes); keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.900, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 20 metres along its route. From 00:00.900 to 00:03.167, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 48 metres along its route, drops about a metre; the vehicle moves about 48 metres along its route, drops about a metre; the vehicle moves about 49 metres along its route, drops about a metre. From 00:03.167 to 00:03.533, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 7 metres along its route, drops a short distance; the vehicle moves about 7 metres along its route, drops a short distance; the vehicle moves about 7 metres along its route, drops a short distance. From 00:03.533 to 00:04.700, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 24 metres along its route, drops about one and a half metres; the vehicle moves about 24 metres along its route, drops about one and a half metres; the vehicle moves about 25 metres along its route, drops about one and a half metres. From 00:04.700 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-068-sky-plane/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-068-sky-plane/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e2010684836966a7bb0b6a089e3e507086b94ac6
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-068-sky-plane/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a matte charcoal fuselage, polished aluminium wings, and clear cockpit glazing, flying through the air, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage, polished aluminium wings, and clear cockpit glazing, maintaining a steady lateral separation from the main aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a matte charcoal fuselage, polished aluminium wings, and clear cockpit glazing, flying through the air, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage, polished aluminium wings, and clear cockpit glazing, maintaining a steady lateral separation from the main aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing confidently in the cockpit of the aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes); keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.900, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 20 metres along its route. From 00:00.900 to 00:03.167, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 48 metres along its route, drops about a metre; the vehicle moves about 48 metres along its route, drops about a metre; the vehicle moves about 49 metres along its route, drops about a metre. From 00:03.167 to 00:03.533, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 7 metres along its route, drops a short distance; the vehicle moves about 7 metres along its route, drops a short distance; the vehicle moves about 7 metres along its route, drops a short distance. From 00:03.533 to 00:04.700, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 24 metres along its route, drops about one and a half metres; the vehicle moves about 24 metres along its route, drops about one and a half metres; the vehicle moves about 25 metres along its route, drops about one and a half metres. From 00:04.700 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-068-sky-plane/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-068-sky-plane/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..cdfdef029f508776e4540bc41a157ac1df9642e9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-068-sky-plane/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a matte charcoal fuselage, polished aluminium wings, and clear cockpit glazing, flying through the air, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage, polished aluminium wings, and clear cockpit glazing, maintaining a steady lateral separation from the main aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a matte charcoal fuselage, polished aluminium wings, and clear cockpit glazing, flying through the air, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage, polished aluminium wings, and clear cockpit glazing, maintaining a steady lateral separation from the main aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing confidently in the cockpit of the aircraft, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes); keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.900, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 20 metres along its route. From 00:00.900 to 00:03.167, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 48 metres along its route, drops about a metre; the vehicle moves about 48 metres along its route, drops about a metre; the vehicle moves about 49 metres along its route, drops about a metre. From 00:03.167 to 00:03.533, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 7 metres along its route, drops a short distance; the vehicle moves about 7 metres along its route, drops a short distance; the vehicle moves about 7 metres along its route, drops a short distance. From 00:03.533 to 00:04.700, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 24 metres along its route, drops about one and a half metres; the vehicle moves about 24 metres along its route, drops about one and a half metres; the vehicle moves about 25 metres along its route, drops about one and a half metres. From 00:04.700 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre; the vehicle moves about 6 metres along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, late-afternoon light through drifting cloud, thin high cirrus, graded washes with almost no visible strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-069-sky-plane/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-069-sky-plane/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..06348983ef82ca954078e8cd712a003160799abd
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-069-sky-plane/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object. <Subject 1> is not in frame at 00:00.000 and enters at about 0.9 s.
+<Subject 1> is the protagonist: not in frame at 00:00.000; it enters the frame at about 0.9 s; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished cream fuselage, dark slate wings, and clear cockpit glazing, flying level through the air - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage and dark grey wings, maintaining a steady flight path in the distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the object proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (enters [Shot 1] at about 0.9 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished cream fuselage, dark slate wings, and clear cockpit glazing, flying level through the air - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage and dark grey wings, maintaining a steady flight path in the distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the open cockpit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.867, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 19 metres along its route. From 00:00.867 to 00:03.200, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 50 metres along its route, drops about half a metre; the vehicle moves about 50 metres along its route, drops about half a metre; the vehicle moves about 52 metres along its route, drops about half a metre. From 00:03.200 to 00:03.533, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route; the vehicle moves about 6 metres along its route; the vehicle moves about 7 metres along its route. From 00:03.533 to 00:04.733, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 25 metres along its route, drops about a metre; the vehicle moves about 25 metres along its route, drops about a metre; the vehicle moves about 26 metres along its route, drops about a metre. From 00:04.733 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 5 metres along its route, drops a short distance; the vehicle moves about 5 metres along its route, drops a short distance; the vehicle moves about 5 metres along its route, drops a short distance. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. The piloted propeller aircraft in the foreground covers the protagonist from about 0.0 s to 0.9 s; the piloted propeller aircraft in the foreground covers the protagonist from about 1.2 s to 5.0 s. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-069-sky-plane/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-069-sky-plane/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6b5a4dbcf0865700965d9fea99cd45c3764e8bc2
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-069-sky-plane/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object. <Subject 1> is not in frame at 00:00.000 and enters at about 0.9 s.
+<Subject 1> is the protagonist: not in frame at 00:00.000; it enters the frame at about 0.9 s; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished cream fuselage, dark slate wings, and clear cockpit glazing, flying level through the air - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage and dark grey wings, maintaining a steady flight path in the distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (enters [Shot 1] at about 0.9 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished cream fuselage, dark slate wings, and clear cockpit glazing, flying level through the air - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage and dark grey wings, maintaining a steady flight path in the distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the open cockpit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.867, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 19 metres along its route. From 00:00.867 to 00:03.200, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 50 metres along its route, drops about half a metre; the vehicle moves about 50 metres along its route, drops about half a metre; the vehicle moves about 52 metres along its route, drops about half a metre. From 00:03.200 to 00:03.533, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route; the vehicle moves about 6 metres along its route; the vehicle moves about 7 metres along its route. From 00:03.533 to 00:04.733, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 25 metres along its route, drops about a metre; the vehicle moves about 25 metres along its route, drops about a metre; the vehicle moves about 26 metres along its route, drops about a metre. From 00:04.733 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 5 metres along its route, drops a short distance; the vehicle moves about 5 metres along its route, drops a short distance; the vehicle moves about 5 metres along its route, drops a short distance. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. The piloted propeller aircraft in the foreground covers the protagonist from about 0.0 s to 0.9 s; the piloted propeller aircraft in the foreground covers the protagonist from about 1.2 s to 5.0 s. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-069-sky-plane/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-069-sky-plane/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a2156ef3cc39cb11da6ef1f5f4317b6e255cb000
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-069-sky-plane/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 object proxies standing for <Subject 3> (the piloted propeller aircraft), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object. <Subject 1> is not in frame at 00:00.000 and enters at about 0.9 s.
+<Subject 1> is the protagonist: not in frame at 00:00.000; it enters the frame at about 0.9 s; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished cream fuselage, dark slate wings, and clear cockpit glazing, flying level through the air - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage and dark grey wings, maintaining a steady flight path in the distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the piloted propeller aircraft; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> flying a propeller aircraft that cruises and banks through turns while a companion aircraft keeps formation, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (enters [Shot 1] at about 0.9 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A single-engine propeller monoplane with a polished cream fuselage, dark slate wings, and clear cockpit glazing, flying level through the air - at the start of the clip its front away from the camera. A single-engine propeller monoplane with a matte slate fuselage and dark grey wings, maintaining a steady flight path in the distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the open cockpit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted propeller aircraft, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.867, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 19 metres along its route. From 00:00.867 to 00:03.200, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 50 metres along its route, drops about half a metre; the vehicle moves about 50 metres along its route, drops about half a metre; the vehicle moves about 52 metres along its route, drops about half a metre. From 00:03.200 to 00:03.533, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 6 metres along its route; the vehicle moves about 6 metres along its route; the vehicle moves about 7 metres along its route. From 00:03.533 to 00:04.733, the protagonist's propeller aircraft banks into a turn, rolling and changing heading and altitude, while the companion aircraft holds formation nearby; the protagonist moves about 25 metres along its route, drops about a metre; the vehicle moves about 25 metres along its route, drops about a metre; the vehicle moves about 26 metres along its route, drops about a metre. From 00:04.733 to 00:05.000, the protagonist's propeller aircraft continues forward through the air while the companion aircraft holds formation nearby; the protagonist moves about 5 metres along its route, drops a short distance; the vehicle moves about 5 metres along its route, drops a short distance; the vehicle moves about 5 metres along its route, drops a short distance. Recorded relations: the protagonist occupies the piloted propeller aircraft from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted propeller aircraft from 00:00.000 to 00:05.000. The piloted propeller aircraft is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. The piloted propeller aircraft in the foreground covers the protagonist from about 0.0 s to 0.9 s; the piloted propeller aircraft in the foreground covers the protagonist from about 1.2 s to 5.0 s. Keep the reference's fixed 55-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..dba3557c5297d886c4fee69f61385e6096eafc41
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/colored-obb/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains no human proxy and exactly 2 object proxies standing for <Subject 3> (the piloted utility helicopter), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A medium transport helicopter with a matte slate-grey fuselage, black rotor blades, and clear cockpit glazing, airborne, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. A twin-engine propeller aircraft with a metallic silver fuselage, fixed wings, and clear cockpit glazing, maintaining lateral separation, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the piloted utility helicopter; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the object proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is vertically ascending in a utility helicopter, flying forward and continuing a controlled descent, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A medium transport helicopter with a matte slate-grey fuselage, black rotor blades, and clear cockpit glazing, airborne, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. A twin-engine propeller aircraft with a metallic silver fuselage, fixed wings, and clear cockpit glazing, maintaining lateral separation, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same impressionist oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted utility helicopter, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.267, the protagonist's helicopter climbs vertically while the companion aircraft holds position nearby; the protagonist rises about 2 metres; the vehicle rises about 2 metres; the vehicle moves about 2 metres along its route, rises about 2 metres. From 00:02.267 to 00:04.133, the protagonist's helicopter flies forward and yaws through the air while the companion aircraft keeps station nearby; the protagonist moves about half a metre along its route, is airborne for about 1.9 seconds; the vehicle moves about half a metre along its route; the vehicle moves about one and a half metres along its route. From 00:04.133 to 00:05.000, the protagonist's helicopter eases off the controls, slows its horizontal motion and continues a controlled descent while the companion aircraft holds position nearby; the protagonist drops about half a metre, is airborne for about 0.9 seconds; the vehicle drops about half a metre; the vehicle moves about half a metre along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted utility helicopter from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted utility helicopter from 00:00.000 to 00:05.000. The piloted utility helicopter is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..cb309ddcd1b94da39322ae058c4cd476ed3a8fd9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/mixed/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains no human proxy and exactly 2 object proxies standing for <Subject 3> (the piloted utility helicopter), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A medium transport helicopter with a matte slate-grey fuselage, black rotor blades, and clear cockpit glazing, airborne, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. A twin-engine propeller aircraft with a metallic silver fuselage, fixed wings, and clear cockpit glazing, maintaining lateral separation, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the piloted utility helicopter; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is vertically ascending in a utility helicopter, flying forward and continuing a controlled descent, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A medium transport helicopter with a matte slate-grey fuselage, black rotor blades, and clear cockpit glazing, airborne, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. A twin-engine propeller aircraft with a metallic silver fuselage, fixed wings, and clear cockpit glazing, maintaining lateral separation, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same impressionist oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted utility helicopter, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.267, the protagonist's helicopter climbs vertically while the companion aircraft holds position nearby; the protagonist rises about 2 metres; the vehicle rises about 2 metres; the vehicle moves about 2 metres along its route, rises about 2 metres. From 00:02.267 to 00:04.133, the protagonist's helicopter flies forward and yaws through the air while the companion aircraft keeps station nearby; the protagonist moves about half a metre along its route, is airborne for about 1.9 seconds; the vehicle moves about half a metre along its route; the vehicle moves about one and a half metres along its route. From 00:04.133 to 00:05.000, the protagonist's helicopter eases off the controls, slows its horizontal motion and continues a controlled descent while the companion aircraft holds position nearby; the protagonist drops about half a metre, is airborne for about 0.9 seconds; the vehicle drops about half a metre; the vehicle moves about half a metre along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted utility helicopter from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted utility helicopter from 00:00.000 to 00:05.000. The piloted utility helicopter is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..529d74503ee61e2557282f6c60083c1c4818e321
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-073-sky-helicopter/first-person/white/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains no human proxy and exactly 2 object proxies standing for <Subject 3> (the piloted utility helicopter), <Subject 4> (the driverless companion aircraft maintaining lateral separation); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A medium transport helicopter with a matte slate-grey fuselage, black rotor blades, and clear cockpit glazing, airborne, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. A twin-engine propeller aircraft with a metallic silver fuselage, fixed wings, and clear cockpit glazing, maintaining lateral separation, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the piloted utility helicopter; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the driverless companion aircraft maintaining lateral separation; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is vertically ascending in a utility helicopter, flying forward and continuing a controlled descent, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: the recorded environment. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A medium transport helicopter with a matte slate-grey fuselage, black rotor blades, and clear cockpit glazing, airborne, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. A twin-engine propeller aircraft with a metallic silver fuselage, fixed wings, and clear cockpit glazing, maintaining lateral separation, painted in the same impressionist oil painting manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same impressionist oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the piloted utility helicopter, realized with the materials above. <Subject 4> is the driverless companion aircraft maintaining lateral separation, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.267, the protagonist's helicopter climbs vertically while the companion aircraft holds position nearby; the protagonist rises about 2 metres; the vehicle rises about 2 metres; the vehicle moves about 2 metres along its route, rises about 2 metres. From 00:02.267 to 00:04.133, the protagonist's helicopter flies forward and yaws through the air while the companion aircraft keeps station nearby; the protagonist moves about half a metre along its route, is airborne for about 1.9 seconds; the vehicle moves about half a metre along its route; the vehicle moves about one and a half metres along its route. From 00:04.133 to 00:05.000, the protagonist's helicopter eases off the controls, slows its horizontal motion and continues a controlled descent while the companion aircraft holds position nearby; the protagonist drops about half a metre, is airborne for about 0.9 seconds; the vehicle drops about half a metre; the vehicle moves about half a metre along its route, drops about half a metre. Recorded relations: the protagonist occupies the piloted utility helicopter from 00:00.000 to 00:05.000; <Subject 4> flies in formation with the piloted utility helicopter from 00:00.000 to 00:05.000. The piloted utility helicopter is one continuous object throughout and ends the clip a short distance from the protagonist. The driverless companion aircraft maintaining lateral separation is one continuous object throughout and ends the clip about 27 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, impressionist oil painting: visible thick impasto brushstrokes, layered opaque pigment, canvas texture, rich colour mixing and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a high-key pastel palette, late-afternoon light through drifting cloud, clear thin air, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bed12ead129e18a086fe15e46e9410e3d82dd0b1
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/colored-obb/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving adult participant 1), <Subject 9> (independently moving adult participant 2), <Subject 10> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the twin-nozzle jetpack), <Subject 5> (the companion jetpack), <Subject 6> (the companion jetpack), <Subject 7> (the companion jetpack); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.4 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack with a sleek, glossy metallic finish in slate grey and amber accents, featuring sharp highlights and a clean, futuristic design, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the twin-nozzle jetpack; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the geometric jet exhaust: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 5> is the companion jetpack; in <Video 1> the solid proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the companion jetpack; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 7> is the companion jetpack; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 8> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+<Subject 10> is independently moving adult participant 3; in <Video 1> the human proxy tagged #3ef736; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> igniting a jetpack, ascending and moving through the air with other flying companions, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack with a sleek, glossy metallic finish in slate grey and amber accents, featuring sharp highlights and a clean, futuristic design, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes). <Subject 1>, the protagonist, has the following target appearance: A field explorer with a muscular, exaggerated heroic build and tiny feet, featuring long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the twin-nozzle jetpack, realized with the materials above. <Subject 4> is the geometric jet exhaust, realized with the materials above. <Subject 5> is the companion jetpack, realized with the materials above. <Subject 6> is the companion jetpack, realized with the materials above. <Subject 7> is the companion jetpack, realized with the materials above. <Subject 8> (independently moving adult participant 1) is a park ranger with a muscular, exaggerated heroic build and tiny feet, featuring short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 2) is a hiker with a muscular, exaggerated heroic build and tiny feet, featuring short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, seen only from behind, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving adult participant 3) is a student with a muscular, exaggerated heroic build and tiny feet, featuring shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.467, the protagonist ignites the jetpack; independently moving adult participant 1 moves a short distance along its route; the jetpack moves a short distance along its route. From 00:00.467 to 00:03.367, the protagonist rises off the ground on the thrust of the jetpack; the protagonist moves about 4 metres along its route, rises about 5 metres; the effect moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 1 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 2 metres along its route, rises about 5 metres; independently moving adult participant 2 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 3 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 2 metres along its route, rises about 5 metres. From 00:03.367 to 00:05.000, the protagonist moves horizontally through the air at a steady height with the other flying companions; the protagonist moves about 5 metres along its route, is airborne for about 1.6 seconds; the effect moves about 5 metres along its route; independently moving adult participant 1 moves about a metre along its route; the jetpack moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route. Recorded relations: the protagonist wears the twin-nozzle jetpack from 00:00.000 to 00:05.000. The twin-nozzle jetpack is one continuous object throughout and ends the clip a short distance from the protagonist. The geometric jet exhaust is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 6 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 1.8 s to 2.0 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.1 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 3 from about 2.4 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..701189345c62ea20ffbf79b15558ca221d46cf4c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/mixed/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving adult participant 1), <Subject 9> (independently moving adult participant 2), <Subject 10> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the twin-nozzle jetpack), <Subject 5> (the companion jetpack), <Subject 6> (the companion jetpack), <Subject 7> (the companion jetpack); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.4 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack with a sleek, glossy metallic finish in slate grey and amber accents, featuring sharp highlights and a clean, futuristic design, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the twin-nozzle jetpack; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the geometric jet exhaust: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 5> is the companion jetpack; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the companion jetpack; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 7> is the companion jetpack; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 8> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 10> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> igniting a jetpack, ascending and moving through the air with other flying companions, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack with a sleek, glossy metallic finish in slate grey and amber accents, featuring sharp highlights and a clean, futuristic design, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes). <Subject 1>, the protagonist, has the following target appearance: A field explorer with a muscular, exaggerated heroic build and tiny feet, featuring long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the twin-nozzle jetpack, realized with the materials above. <Subject 4> is the geometric jet exhaust, realized with the materials above. <Subject 5> is the companion jetpack, realized with the materials above. <Subject 6> is the companion jetpack, realized with the materials above. <Subject 7> is the companion jetpack, realized with the materials above. <Subject 8> (independently moving adult participant 1) is a park ranger with a muscular, exaggerated heroic build and tiny feet, featuring short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 2) is a hiker with a muscular, exaggerated heroic build and tiny feet, featuring short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, seen only from behind, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving adult participant 3) is a student with a muscular, exaggerated heroic build and tiny feet, featuring shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.467, the protagonist ignites the jetpack; independently moving adult participant 1 moves a short distance along its route; the jetpack moves a short distance along its route. From 00:00.467 to 00:03.367, the protagonist rises off the ground on the thrust of the jetpack; the protagonist moves about 4 metres along its route, rises about 5 metres; the effect moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 1 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 2 metres along its route, rises about 5 metres; independently moving adult participant 2 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 3 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 2 metres along its route, rises about 5 metres. From 00:03.367 to 00:05.000, the protagonist moves horizontally through the air at a steady height with the other flying companions; the protagonist moves about 5 metres along its route, is airborne for about 1.6 seconds; the effect moves about 5 metres along its route; independently moving adult participant 1 moves about a metre along its route; the jetpack moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route. Recorded relations: the protagonist wears the twin-nozzle jetpack from 00:00.000 to 00:05.000. The twin-nozzle jetpack is one continuous object throughout and ends the clip a short distance from the protagonist. The geometric jet exhaust is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 6 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 1.8 s to 2.0 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.1 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 3 from about 2.4 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0bd0bbdacf370225e4007875a707e45563174b7d
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-075-sky-jetpack/follow/white/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving adult participant 1), <Subject 9> (independently moving adult participant 2), <Subject 10> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the twin-nozzle jetpack), <Subject 5> (the companion jetpack), <Subject 6> (the companion jetpack), <Subject 7> (the companion jetpack); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.4 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack with a sleek, glossy metallic finish in slate grey and amber accents, featuring sharp highlights and a clean, futuristic design, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the twin-nozzle jetpack; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the geometric jet exhaust: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 5> is the companion jetpack; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the companion jetpack; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the companion jetpack; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 8> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 10> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> igniting a jetpack, ascending and moving through the air with other flying companions, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack with a sleek, glossy metallic finish in slate grey and amber accents, featuring sharp highlights and a clean, futuristic design, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes). <Subject 1>, the protagonist, has the following target appearance: A field explorer with a muscular, exaggerated heroic build and tiny feet, featuring long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the twin-nozzle jetpack, realized with the materials above. <Subject 4> is the geometric jet exhaust, realized with the materials above. <Subject 5> is the companion jetpack, realized with the materials above. <Subject 6> is the companion jetpack, realized with the materials above. <Subject 7> is the companion jetpack, realized with the materials above. <Subject 8> (independently moving adult participant 1) is a park ranger with a muscular, exaggerated heroic build and tiny feet, featuring short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 2) is a hiker with a muscular, exaggerated heroic build and tiny feet, featuring short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, seen only from behind, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving adult participant 3) is a student with a muscular, exaggerated heroic build and tiny feet, featuring shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, drawn as an anime character in the same soft-shaded anime manner (character design: rounded youthful proportions with large expressive eyes) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.467, the protagonist ignites the jetpack; independently moving adult participant 1 moves a short distance along its route; the jetpack moves a short distance along its route. From 00:00.467 to 00:03.367, the protagonist rises off the ground on the thrust of the jetpack; the protagonist moves about 4 metres along its route, rises about 5 metres; the effect moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 1 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 2 metres along its route, rises about 5 metres; independently moving adult participant 2 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 3 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 2 metres along its route, rises about 5 metres. From 00:03.367 to 00:05.000, the protagonist moves horizontally through the air at a steady height with the other flying companions; the protagonist moves about 5 metres along its route, is airborne for about 1.6 seconds; the effect moves about 5 metres along its route; independently moving adult participant 1 moves about a metre along its route; the jetpack moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route. Recorded relations: the protagonist wears the twin-nozzle jetpack from 00:00.000 to 00:05.000. The twin-nozzle jetpack is one continuous object throughout and ends the clip a short distance from the protagonist. The geometric jet exhaust is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 9 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 6 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 1.8 s to 2.0 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.1 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 3 from about 2.4 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — soft-shaded anime: gently shaded characters, thin contour lines, lush hand-painted backgrounds and warm natural light; a sun-bleached faded palette, misty morning backlight, light drifting haze; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e0d60b616095fa421a23e654c9171195d75c43ee
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/colored-obb/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving adult participant 1), <Subject 9> (independently moving adult participant 2), <Subject 10> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the twin-nozzle jetpack), <Subject 5> (the companion jetpack), <Subject 6> (the companion jetpack), <Subject 7> (the companion jetpack); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.4 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack constructed of polished brass and riveted iron, featuring intricate copper piping and glowing amber gauges, worn on the back of the protagonist, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the twin-nozzle jetpack; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the geometric jet exhaust: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 5> is the companion jetpack; in <Video 1> the solid proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the companion jetpack; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 7> is the companion jetpack; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 8> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+<Subject 10> is independently moving adult participant 3; in <Video 1> the human proxy tagged #3ef736; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> igniting a jetpack, ascending and moving through the air with other flying companions, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack constructed of polished brass and riveted iron, featuring intricate copper piping and glowing amber gauges, worn on the back of the protagonist, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs). <Subject 1>, the protagonist, has the following target appearance: An adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing centrally in the foreground with a confident posture, seen only from behind, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the twin-nozzle jetpack, realized with the materials above. <Subject 4> is the geometric jet exhaust, realized with the materials above. <Subject 5> is the companion jetpack, realized with the materials above. <Subject 6> is the companion jetpack, realized with the materials above. <Subject 7> is the companion jetpack, realized with the materials above. <Subject 8> (independently moving adult participant 1) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and leather gloves, standing to the left with a practical, ready stance, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 2) is an adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing slightly behind and to the right with a composed demeanor, seen only from behind, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving adult participant 3) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the left rear with a relaxed, observant posture, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.467, the protagonist ignites the jetpack. From 00:00.467 to 00:03.367, the protagonist rises off the ground on the thrust of the jetpack; the protagonist moves about 4 metres along its route, rises about 5 metres; the effect moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 1 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 3 metres along its route, rises about 5 metres; independently moving adult participant 2 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 3 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 3 metres along its route, rises about 5 metres. From 00:03.367 to 00:05.000, the protagonist moves horizontally through the air at a steady height with the other flying companions; the protagonist moves about 5 metres along its route, is airborne for about 1.6 seconds; the effect moves about 5 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; the jetpack moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route. Recorded relations: the protagonist wears the twin-nozzle jetpack from 00:00.000 to 00:05.000. The twin-nozzle jetpack is one continuous object throughout and ends the clip a short distance from the protagonist. The geometric jet exhaust is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 10 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about a metre from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 7 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 1.0 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.2 s to 3.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5244083b581c18c8568951df0c9fa411afb3816c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/mixed/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving adult participant 1), <Subject 9> (independently moving adult participant 2), <Subject 10> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the twin-nozzle jetpack), <Subject 5> (the companion jetpack), <Subject 6> (the companion jetpack), <Subject 7> (the companion jetpack); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.4 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack constructed of polished brass and riveted iron, featuring intricate copper piping and glowing amber gauges, worn on the back of the protagonist, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the twin-nozzle jetpack; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the geometric jet exhaust: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 5> is the companion jetpack; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the companion jetpack; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 7> is the companion jetpack; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 8> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 10> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> igniting a jetpack, ascending and moving through the air with other flying companions, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack constructed of polished brass and riveted iron, featuring intricate copper piping and glowing amber gauges, worn on the back of the protagonist, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs). <Subject 1>, the protagonist, has the following target appearance: An adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing centrally in the foreground with a confident posture, seen only from behind, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the twin-nozzle jetpack, realized with the materials above. <Subject 4> is the geometric jet exhaust, realized with the materials above. <Subject 5> is the companion jetpack, realized with the materials above. <Subject 6> is the companion jetpack, realized with the materials above. <Subject 7> is the companion jetpack, realized with the materials above. <Subject 8> (independently moving adult participant 1) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and leather gloves, standing to the left with a practical, ready stance, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 2) is an adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing slightly behind and to the right with a composed demeanor, seen only from behind, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving adult participant 3) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the left rear with a relaxed, observant posture, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.467, the protagonist ignites the jetpack. From 00:00.467 to 00:03.367, the protagonist rises off the ground on the thrust of the jetpack; the protagonist moves about 4 metres along its route, rises about 5 metres; the effect moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 1 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 3 metres along its route, rises about 5 metres; independently moving adult participant 2 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 3 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 3 metres along its route, rises about 5 metres. From 00:03.367 to 00:05.000, the protagonist moves horizontally through the air at a steady height with the other flying companions; the protagonist moves about 5 metres along its route, is airborne for about 1.6 seconds; the effect moves about 5 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; the jetpack moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route. Recorded relations: the protagonist wears the twin-nozzle jetpack from 00:00.000 to 00:05.000. The twin-nozzle jetpack is one continuous object throughout and ends the clip a short distance from the protagonist. The geometric jet exhaust is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 10 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about a metre from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 7 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 1.0 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.2 s to 3.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a2583f949c51ef3c30125773910673b0ff75fc9e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-076-sky-jetpack/follow/white/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 8> (independently moving adult participant 1), <Subject 9> (independently moving adult participant 2), <Subject 10> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the twin-nozzle jetpack), <Subject 5> (the companion jetpack), <Subject 6> (the companion jetpack), <Subject 7> (the companion jetpack); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.4 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack constructed of polished brass and riveted iron, featuring intricate copper piping and glowing amber gauges, worn on the back of the protagonist, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the twin-nozzle jetpack; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the geometric jet exhaust: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 5> is the companion jetpack; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the companion jetpack; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the companion jetpack; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 8> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 10> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> igniting a jetpack, ascending and moving through the air with other flying companions, involving <Subject 3> to <Subject 10>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A twin-nozzle jetpack constructed of polished brass and riveted iron, featuring intricate copper piping and glowing amber gauges, worn on the back of the protagonist, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs). <Subject 1>, the protagonist, has the following target appearance: An adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing centrally in the foreground with a confident posture, seen only from behind, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the twin-nozzle jetpack, realized with the materials above. <Subject 4> is the geometric jet exhaust, realized with the materials above. <Subject 5> is the companion jetpack, realized with the materials above. <Subject 6> is the companion jetpack, realized with the materials above. <Subject 7> is the companion jetpack, realized with the materials above. <Subject 8> (independently moving adult participant 1) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and leather gloves, standing to the left with a practical, ready stance, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 2) is an adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing slightly behind and to the right with a composed demeanor, seen only from behind, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 10> (independently moving adult participant 3) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the left rear with a relaxed, observant posture, drawn as an anime character in the same modern digital anime manner (character design: tall slender figures with small faces and long limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.467, the protagonist ignites the jetpack. From 00:00.467 to 00:03.367, the protagonist rises off the ground on the thrust of the jetpack; the protagonist moves about 4 metres along its route, rises about 5 metres; the effect moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 1 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 3 metres along its route, rises about 5 metres; independently moving adult participant 2 moves about 2 metres along its route, rises about 5 metres; the jetpack moves about 4 metres along its route, rises about 5 metres; independently moving adult participant 3 moves about 3 metres along its route, rises about 5 metres; the jetpack moves about 3 metres along its route, rises about 5 metres. From 00:03.367 to 00:05.000, the protagonist moves horizontally through the air at a steady height with the other flying companions; the protagonist moves about 5 metres along its route, is airborne for about 1.6 seconds; the effect moves about 5 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; the jetpack moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the jetpack moves about one and a half metres along its route. Recorded relations: the protagonist wears the twin-nozzle jetpack from 00:00.000 to 00:05.000. The twin-nozzle jetpack is one continuous object throughout and ends the clip a short distance from the protagonist. The geometric jet exhaust is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 10 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about a metre from the protagonist. <Subject 7> is one continuous object throughout and ends the clip about 7 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 1.0 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 2 from about 2.2 s to 3.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — modern digital anime: gradient shading, glossy highlights, clean thin contour lines, vivid high-contrast colours; a monochrome-leaning palette with one accent colour, flat overcast daylight, scattered cumulus clouds, backgrounds softened as if slightly out of focus; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..635b5dc342bf4fe6813fe87114550b90e099f5a7
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 human proxies standing for <Subject 6> (the seated driver of rival racing car 1), <Subject 7> (the seated driver of rival racing car 2) and 3 object proxies standing for <Subject 3> (the protagonist's racing sedan with a seated driver), <Subject 4> (the rival sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #a7f736 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A coastal beach with sloping sand, large shoreline rocks, a pier and small boats. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy pearl white with racing decals, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, finished in metallic slate grey, appearing as a complete vehicle with a sleek, aerodynamic profile. A rally hatchback with mud flaps, finished in matte rust orange, featuring a rugged body kit and roof rack - at the start of the clip its front away from the camera. A large, heavy steel gantry structure painted in matte charcoal grey, spanning the width of the track with a weathered, industrial finish. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the protagonist's racing sedan with a seated driver; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 4> is the rival sedan with a seated driver; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the seated driver of rival racing car 1; in <Video 1> the human proxy tagged #3ef736; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 2; in <Video 1> the human proxy tagged #f73677 that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A coastal beach with sloping sand, large shoreline rocks, a pier and small boats. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy pearl white with racing decals, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, finished in metallic slate grey, appearing as a complete vehicle with a sleek, aerodynamic profile. A rally hatchback with mud flaps, finished in matte rust orange, featuring a rugged body kit and roof rack - at the start of the clip its front away from the camera. A large, heavy steel gantry structure painted in matte charcoal grey, spanning the width of the track with a weathered, industrial finish. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 4> is the rival sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> (the seated driver of rival racing car 1) is a realistic adult road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival-1 vehicle, and keeps exactly this appearance. <Subject 7> (the seated driver of rival racing car 2) is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes and charcoal uniform trousers, seated inside the rival-2 vehicle, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.633, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 59 metres along its route; the vehicle moves about 59 metres along its route; the vehicle moves about 60 metres along its route; the vehicle moves about 87 metres along its route; the seated driver of rival racing car 1 moves about 60 metres along its route; the seated driver of rival racing car 2 moves about 87 metres along its route. From 00:02.633 to 00:04.400, the protagonist's sedan pulls out alongside a rival sedan and tries to pass while the other rival sedan keeps racing on its own line; the protagonist moves about 46 metres along its route; the vehicle moves about 46 metres along its route; the vehicle moves about 41 metres along its route; the vehicle moves about 60 metres along its route; the seated driver of rival racing car 1 moves about 41 metres along its route; the seated driver of rival racing car 2 moves about 60 metres along its route. From 00:04.400 to 00:05.000, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 14 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 14 metres along its route; the seated driver of rival racing car 2 moves about 18 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 4> from 00:00.000 to 00:05.000; the seated driver of rival racing car 2 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 3> follows <Subject 4> from 00:00.000 to 00:02.633; <Subject 3> overtakes <Subject 4> from 00:01.267 to 00:05.000; <Subject 3> follows <Subject 5> from 00:04.400 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 49 metres from the protagonist. The protagonist's racing sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.4 s; the protagonist's racing sedan with a seated driver in the foreground covers the seated driver of rival racing car 2 from about 0.7 s to 5.0 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 2 from about 0.7 s to 5.0 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b05bd61b0fbe6bfe8e5d464796a67a70f70b76c9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 blue mannequin proxies standing for <Subject 6> (the seated driver of rival racing car 1), <Subject 7> (the seated driver of rival racing car 2) and 3 object proxies standing for <Subject 3> (the protagonist's racing sedan with a seated driver), <Subject 4> (the rival sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A coastal beach with sloping sand, large shoreline rocks, a pier and small boats. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy pearl white with racing decals, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, finished in metallic slate grey, appearing as a complete vehicle with a sleek, aerodynamic profile. A rally hatchback with mud flaps, finished in matte rust orange, featuring a rugged body kit and roof rack - at the start of the clip its front away from the camera. A large, heavy steel gantry structure painted in matte charcoal grey, spanning the width of the track with a weathered, industrial finish. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the protagonist's racing sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the seated driver of rival racing car 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A coastal beach with sloping sand, large shoreline rocks, a pier and small boats. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy pearl white with racing decals, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, finished in metallic slate grey, appearing as a complete vehicle with a sleek, aerodynamic profile. A rally hatchback with mud flaps, finished in matte rust orange, featuring a rugged body kit and roof rack - at the start of the clip its front away from the camera. A large, heavy steel gantry structure painted in matte charcoal grey, spanning the width of the track with a weathered, industrial finish. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 4> is the rival sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> (the seated driver of rival racing car 1) is a realistic adult road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival-1 vehicle, and keeps exactly this appearance. <Subject 7> (the seated driver of rival racing car 2) is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes and charcoal uniform trousers, seated inside the rival-2 vehicle, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.633, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 59 metres along its route; the vehicle moves about 59 metres along its route; the vehicle moves about 60 metres along its route; the vehicle moves about 87 metres along its route; the seated driver of rival racing car 1 moves about 60 metres along its route; the seated driver of rival racing car 2 moves about 87 metres along its route. From 00:02.633 to 00:04.400, the protagonist's sedan pulls out alongside a rival sedan and tries to pass while the other rival sedan keeps racing on its own line; the protagonist moves about 46 metres along its route; the vehicle moves about 46 metres along its route; the vehicle moves about 41 metres along its route; the vehicle moves about 60 metres along its route; the seated driver of rival racing car 1 moves about 41 metres along its route; the seated driver of rival racing car 2 moves about 60 metres along its route. From 00:04.400 to 00:05.000, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 14 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 14 metres along its route; the seated driver of rival racing car 2 moves about 18 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 4> from 00:00.000 to 00:05.000; the seated driver of rival racing car 2 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 3> follows <Subject 4> from 00:00.000 to 00:02.633; <Subject 3> overtakes <Subject 4> from 00:01.267 to 00:05.000; <Subject 3> follows <Subject 5> from 00:04.400 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 49 metres from the protagonist. The protagonist's racing sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.4 s; the protagonist's racing sedan with a seated driver in the foreground covers the seated driver of rival racing car 2 from about 0.7 s to 5.0 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 2 from about 0.7 s to 5.0 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6eee404fe61b2fca884300aef905bf2e0357c430
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-078-racing-racing/first-person/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 grey block-figure proxies standing for <Subject 6> (the seated driver of rival racing car 1), <Subject 7> (the seated driver of rival racing car 2) and 3 object proxies standing for <Subject 3> (the protagonist's racing sedan with a seated driver), <Subject 4> (the rival sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A coastal beach with sloping sand, large shoreline rocks, a pier and small boats. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy pearl white with racing decals, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, finished in metallic slate grey, appearing as a complete vehicle with a sleek, aerodynamic profile. A rally hatchback with mud flaps, finished in matte rust orange, featuring a rugged body kit and roof rack - at the start of the clip its front away from the camera. A large, heavy steel gantry structure painted in matte charcoal grey, spanning the width of the track with a weathered, industrial finish. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the protagonist's racing sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the seated driver of rival racing car 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A coastal beach with sloping sand, large shoreline rocks, a pier and small boats. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy pearl white with racing decals, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, finished in metallic slate grey, appearing as a complete vehicle with a sleek, aerodynamic profile. A rally hatchback with mud flaps, finished in matte rust orange, featuring a rugged body kit and roof rack - at the start of the clip its front away from the camera. A large, heavy steel gantry structure painted in matte charcoal grey, spanning the width of the track with a weathered, industrial finish. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 4> is the rival sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> (the seated driver of rival racing car 1) is a realistic adult road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival-1 vehicle, and keeps exactly this appearance. <Subject 7> (the seated driver of rival racing car 2) is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes and charcoal uniform trousers, seated inside the rival-2 vehicle, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.633, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 59 metres along its route; the vehicle moves about 59 metres along its route; the vehicle moves about 60 metres along its route; the vehicle moves about 87 metres along its route; the seated driver of rival racing car 1 moves about 60 metres along its route; the seated driver of rival racing car 2 moves about 87 metres along its route. From 00:02.633 to 00:04.400, the protagonist's sedan pulls out alongside a rival sedan and tries to pass while the other rival sedan keeps racing on its own line; the protagonist moves about 46 metres along its route; the vehicle moves about 46 metres along its route; the vehicle moves about 41 metres along its route; the vehicle moves about 60 metres along its route; the seated driver of rival racing car 1 moves about 41 metres along its route; the seated driver of rival racing car 2 moves about 60 metres along its route. From 00:04.400 to 00:05.000, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 14 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 14 metres along its route; the seated driver of rival racing car 2 moves about 18 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 4> from 00:00.000 to 00:05.000; the seated driver of rival racing car 2 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 3> follows <Subject 4> from 00:00.000 to 00:02.633; <Subject 3> overtakes <Subject 4> from 00:01.267 to 00:05.000; <Subject 3> follows <Subject 5> from 00:04.400 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 14 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 49 metres from the protagonist. The protagonist's racing sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.4 s; the protagonist's racing sedan with a seated driver in the foreground covers the seated driver of rival racing car 2 from about 0.7 s to 5.0 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 2 from about 0.7 s to 5.0 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-080-racing-racing/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-080-racing-racing/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..403f94170019683bc0613ab8efa82a0488a67c2e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-080-racing-racing/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 human proxies standing for <Subject 1> (the protagonist), <Subject 6> (the seated driver of rival racing car 3) and 3 object proxies standing for <Subject 3> (the protagonist's racing sedan with a seated driver), <Subject 4> (the rival sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. The proxy tagged #3ef736 (person) never enters the frame of this shot: no figure or object appears for it and none must be added.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The protagonist's racing sedan is a low racing sedan with a rear wing, painted in a matte charcoal finish with exposed mechanical details and glowing blue underglow, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes) - at the start of the clip its front away from the camera. The rival sedan is a touring-car style racing saloon, finished in a glossy slate grey with racing decals and a prominent rear spoiler, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). The rival sedan is a rally hatchback with mud flaps, painted in a rugged olive green with reinforced bumpers and off-road tires, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the protagonist's racing sedan with a seated driver; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 4> is the rival sedan with a seated driver; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the seated driver of rival racing car 3; in <Video 1> the human proxy tagged #36aff7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The protagonist's racing sedan is a low racing sedan with a rear wing, painted in a matte charcoal finish with exposed mechanical details and glowing blue underglow, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes) - at the start of the clip its front away from the camera. The rival sedan is a touring-car style racing saloon, finished in a glossy slate grey with racing decals and a prominent rear spoiler, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). The rival sedan is a rally hatchback with mud flaps, painted in a rugged olive green with reinforced bumpers and off-road tires, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). <Subject 1>, the protagonist, has the following target appearance: The protagonist is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and practical tools in a belt pouch, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes); keep this identity and outfit consistent throughout the shot. <Subject 3> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 4> is the rival sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> (the seated driver of rival racing car 3) is the seated driver of rival racing car 3 is an adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes), and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 55 metres along its route; the vehicle moves about 55 metres along its route; the vehicle moves about 53 metres along its route; the vehicle moves about 50 metres along its route; the seated driver of rival racing car 3 moves about 50 metres along its route. From 00:02.500 to 00:05.000, the protagonist's sedan keeps following the rival sedan ahead along the track while the other rival sedan keeps racing on its own line; the protagonist moves about 42 metres along its route; the vehicle moves about 41 metres along its route; the vehicle moves about 46 metres along its route; the vehicle moves about 50 metres along its route; the seated driver of rival racing car 3 moves about 50 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 3> follows <Subject 5> from 00:00.000 to 00:01.633; <Subject 3> overtakes <Subject 5> from 00:00.267 to 00:05.000; <Subject 3> follows <Subject 4> from 00:01.700 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 26 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 34 metres from the protagonist. The rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 3 from about 1.5 s to 1.8 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-080-racing-racing/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-080-racing-racing/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d87ff756fb18dafb348b7c2eab11e366d1e375de
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-080-racing-racing/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 6> (the seated driver of rival racing car 3) and 3 object proxies standing for <Subject 3> (the protagonist's racing sedan with a seated driver), <Subject 4> (the rival sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The protagonist's racing sedan is a low racing sedan with a rear wing, painted in a matte charcoal finish with exposed mechanical details and glowing blue underglow, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes) - at the start of the clip its front away from the camera. The rival sedan is a touring-car style racing saloon, finished in a glossy slate grey with racing decals and a prominent rear spoiler, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). The rival sedan is a rally hatchback with mud flaps, painted in a rugged olive green with reinforced bumpers and off-road tires, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the protagonist's racing sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 6> is the seated driver of rival racing car 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The protagonist's racing sedan is a low racing sedan with a rear wing, painted in a matte charcoal finish with exposed mechanical details and glowing blue underglow, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes) - at the start of the clip its front away from the camera. The rival sedan is a touring-car style racing saloon, finished in a glossy slate grey with racing decals and a prominent rear spoiler, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). The rival sedan is a rally hatchback with mud flaps, painted in a rugged olive green with reinforced bumpers and off-road tires, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). <Subject 1>, the protagonist, has the following target appearance: The protagonist is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and practical tools in a belt pouch, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes); keep this identity and outfit consistent throughout the shot. <Subject 3> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 4> is the rival sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> (the seated driver of rival racing car 3) is the seated driver of rival racing car 3 is an adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes), and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 55 metres along its route; the vehicle moves about 55 metres along its route; the vehicle moves about 53 metres along its route; the vehicle moves about 50 metres along its route; the seated driver of rival racing car 3 moves about 50 metres along its route. From 00:02.500 to 00:05.000, the protagonist's sedan keeps following the rival sedan ahead along the track while the other rival sedan keeps racing on its own line; the protagonist moves about 42 metres along its route; the vehicle moves about 41 metres along its route; the vehicle moves about 46 metres along its route; the vehicle moves about 50 metres along its route; the seated driver of rival racing car 3 moves about 50 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 3> follows <Subject 5> from 00:00.000 to 00:01.633; <Subject 3> overtakes <Subject 5> from 00:00.267 to 00:05.000; <Subject 3> follows <Subject 4> from 00:01.700 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 26 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 34 metres from the protagonist. The rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 3 from about 1.5 s to 1.8 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-080-racing-racing/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-080-racing-racing/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f96c8be1dabb8c91f4bf06037358ecf49c40ea32
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-080-racing-racing/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 6> (the seated driver of rival racing car 3) and 3 object proxies standing for <Subject 3> (the protagonist's racing sedan with a seated driver), <Subject 4> (the rival sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The protagonist's racing sedan is a low racing sedan with a rear wing, painted in a matte charcoal finish with exposed mechanical details and glowing blue underglow, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes) - at the start of the clip its front away from the camera. The rival sedan is a touring-car style racing saloon, finished in a glossy slate grey with racing decals and a prominent rear spoiler, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). The rival sedan is a rally hatchback with mud flaps, painted in a rugged olive green with reinforced bumpers and off-road tires, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the protagonist's racing sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the seated driver of rival racing car 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: The protagonist's racing sedan is a low racing sedan with a rear wing, painted in a matte charcoal finish with exposed mechanical details and glowing blue underglow, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes) - at the start of the clip its front away from the camera. The rival sedan is a touring-car style racing saloon, finished in a glossy slate grey with racing decals and a prominent rear spoiler, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). The rival sedan is a rally hatchback with mud flaps, painted in a rugged olive green with reinforced bumpers and off-road tires, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes). <Subject 1>, the protagonist, has the following target appearance: The protagonist is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and practical tools in a belt pouch, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes); keep this identity and outfit consistent throughout the shot. <Subject 3> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 4> is the rival sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> (the seated driver of rival racing car 3) is the seated driver of rival racing car 3 is an adult field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, drawn as an anime character in the same pastel slice-of-life anime manner (character design: angular sharp-jawed figures with narrow eyes), and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 55 metres along its route; the vehicle moves about 55 metres along its route; the vehicle moves about 53 metres along its route; the vehicle moves about 50 metres along its route; the seated driver of rival racing car 3 moves about 50 metres along its route. From 00:02.500 to 00:05.000, the protagonist's sedan keeps following the rival sedan ahead along the track while the other rival sedan keeps racing on its own line; the protagonist moves about 42 metres along its route; the vehicle moves about 41 metres along its route; the vehicle moves about 46 metres along its route; the vehicle moves about 50 metres along its route; the seated driver of rival racing car 3 moves about 50 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 3> follows <Subject 5> from 00:00.000 to 00:01.633; <Subject 3> overtakes <Subject 5> from 00:00.267 to 00:05.000; <Subject 3> follows <Subject 4> from 00:01.700 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 26 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 34 metres from the protagonist. The rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 3 from about 1.5 s to 1.8 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — pastel slice-of-life anime look: thin clean lines, bright airy pastel colours, soft bloom on highlights, tidy detailed everyday backgrounds; a low-saturation earthy palette, golden-hour side light with long soft shadows, still humid air after rain; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-081-racing-racing/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-081-racing-racing/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4651218d4224191c36c2f362a7f0ea438c736fa7
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-081-racing-racing/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 4.2 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a matte charcoal shade, featuring a complete body with doors, windows, and wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, painted in a deep slate blue, with a complete body including doors, windows, and wheels. A low racing sedan with a rear wing, painted in a dark gunmetal grey, with a complete body including doors, windows, and wheels. A massive steel gantry structure with heavy riveted beams and a weathered finish, spanning the width of the yard. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1: not in frame at 00:00.000; it enters the frame at about 4.2 s; in <Video 1> the human proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the human proxy tagged #36aff7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 4.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a matte charcoal shade, featuring a complete body with doors, windows, and wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, painted in a deep slate blue, with a complete body including doors, windows, and wheels. A low racing sedan with a rear wing, painted in a dark gunmetal grey, with a complete body including doors, windows, and wheels. A massive steel gantry structure with heavy riveted beams and a weathered finish, spanning the width of the yard. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is not in frame at 00:00.000 and enters the frame at about 4.2 s; from then on it is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.333, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 30 metres along its route; the vehicle moves about 30 metres along its route; the vehicle moves about 25 metres along its route; the vehicle moves about 43 metres along its route; the seated driver of rival racing car 1 moves about 25 metres along its route; the seated driver of rival racing car 3 moves about 43 metres along its route. From 00:01.333 to 00:04.433, the protagonist's sedan pulls out and passes a rival sedan while the other rival sedan keeps racing on its own line; the protagonist moves about 87 metres along its route; the vehicle moves about 87 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 107 metres along its route; the seated driver of rival racing car 1 moves about 61 metres along its route; the seated driver of rival racing car 3 moves about 107 metres along its route. From 00:04.433 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 20 metres along its route; the vehicle moves about 20 metres along its route; the vehicle moves about 11 metres along its route; the vehicle moves about 19 metres along its route; the seated driver of rival racing car 1 moves about 11 metres along its route; the seated driver of rival racing car 3 moves about 19 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 5> from 00:00.000 to 00:01.333; <Subject 4> overtakes <Subject 5> from 00:00.000 to 00:04.433; <Subject 5> follows the protagonist's racing sedan with a seated driver from 00:04.933 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 18 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 100 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-081-racing-racing/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-081-racing-racing/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..89e0979dba5e96066a582b230766721f34e5a363
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-081-racing-racing/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 4.2 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a matte charcoal shade, featuring a complete body with doors, windows, and wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, painted in a deep slate blue, with a complete body including doors, windows, and wheels. A low racing sedan with a rear wing, painted in a dark gunmetal grey, with a complete body including doors, windows, and wheels. A massive steel gantry structure with heavy riveted beams and a weathered finish, spanning the width of the yard. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the black solid checkpoint proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1: not in frame at 00:00.000; it enters the frame at about 4.2 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 4.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a matte charcoal shade, featuring a complete body with doors, windows, and wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, painted in a deep slate blue, with a complete body including doors, windows, and wheels. A low racing sedan with a rear wing, painted in a dark gunmetal grey, with a complete body including doors, windows, and wheels. A massive steel gantry structure with heavy riveted beams and a weathered finish, spanning the width of the yard. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is not in frame at 00:00.000 and enters the frame at about 4.2 s; from then on it is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.333, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 30 metres along its route; the vehicle moves about 30 metres along its route; the vehicle moves about 25 metres along its route; the vehicle moves about 43 metres along its route; the seated driver of rival racing car 1 moves about 25 metres along its route; the seated driver of rival racing car 3 moves about 43 metres along its route. From 00:01.333 to 00:04.433, the protagonist's sedan pulls out and passes a rival sedan while the other rival sedan keeps racing on its own line; the protagonist moves about 87 metres along its route; the vehicle moves about 87 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 107 metres along its route; the seated driver of rival racing car 1 moves about 61 metres along its route; the seated driver of rival racing car 3 moves about 107 metres along its route. From 00:04.433 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 20 metres along its route; the vehicle moves about 20 metres along its route; the vehicle moves about 11 metres along its route; the vehicle moves about 19 metres along its route; the seated driver of rival racing car 1 moves about 11 metres along its route; the seated driver of rival racing car 3 moves about 19 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 5> from 00:00.000 to 00:01.333; <Subject 4> overtakes <Subject 5> from 00:00.000 to 00:04.433; <Subject 5> follows the protagonist's racing sedan with a seated driver from 00:04.933 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 18 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 100 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-081-racing-racing/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-081-racing-racing/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3c1044a758946a645ebc49e1882e0911a5979a98
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-081-racing-racing/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 4.2 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a matte charcoal shade, featuring a complete body with doors, windows, and wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, painted in a deep slate blue, with a complete body including doors, windows, and wheels. A low racing sedan with a rear wing, painted in a dark gunmetal grey, with a complete body including doors, windows, and wheels. A massive steel gantry structure with heavy riveted beams and a weathered finish, spanning the width of the yard. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1: not in frame at 00:00.000; it enters the frame at about 4.2 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 4.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a matte charcoal shade, featuring a complete body with doors, windows, and wheels - at the start of the clip its front away from the camera. A touring-car style racing saloon, painted in a deep slate blue, with a complete body including doors, windows, and wheels. A low racing sedan with a rear wing, painted in a dark gunmetal grey, with a complete body including doors, windows, and wheels. A massive steel gantry structure with heavy riveted beams and a weathered finish, spanning the width of the yard. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is not in frame at 00:00.000 and enters the frame at about 4.2 s; from then on it is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.333, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 30 metres along its route; the vehicle moves about 30 metres along its route; the vehicle moves about 25 metres along its route; the vehicle moves about 43 metres along its route; the seated driver of rival racing car 1 moves about 25 metres along its route; the seated driver of rival racing car 3 moves about 43 metres along its route. From 00:01.333 to 00:04.433, the protagonist's sedan pulls out and passes a rival sedan while the other rival sedan keeps racing on its own line; the protagonist moves about 87 metres along its route; the vehicle moves about 87 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 107 metres along its route; the seated driver of rival racing car 1 moves about 61 metres along its route; the seated driver of rival racing car 3 moves about 107 metres along its route. From 00:04.433 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 20 metres along its route; the vehicle moves about 20 metres along its route; the vehicle moves about 11 metres along its route; the vehicle moves about 19 metres along its route; the seated driver of rival racing car 1 moves about 11 metres along its route; the seated driver of rival racing car 3 moves about 19 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 5> from 00:00.000 to 00:01.333; <Subject 4> overtakes <Subject 5> from 00:00.000 to 00:04.433; <Subject 5> follows the protagonist's racing sedan with a seated driver from 00:04.933 to 00:05.000. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 18 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 100 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-082-racing-racing/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-082-racing-racing/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..91d6f1cfa14735270758919cdc17994380cf0389
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-082-racing-racing/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 human proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 4.4 s. The proxy tagged #6636f7 (vehicle) never enters the frame of this shot: no figure or object appears for it and none must be added.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A touring-car style racing saloon with a glossy, dark charcoal body, featuring aerodynamic curves, racing decals, and a visible driver inside, positioned in the lower centre of the frame - at the start of the clip its front away from the camera. A touring-car style racing saloon with a glossy, dark charcoal body, similar to the camera-owner's car, positioned further back in the middle right of the frame - at the start of the clip its front toward the screen-right. A touring-car style racing saloon with a glossy, dark charcoal body, identical to the other racing sedans, currently behind the camera and not visible in the frame. A large, dark timber gantry structure with a weathered finish, spanning the width of the track, its beams casting long shadows on the grass below. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the object proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver: not in frame at 00:00.000; it enters the frame at about 4.4 s; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the human proxy tagged #3ef736 that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 4.4 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A touring-car style racing saloon with a glossy, dark charcoal body, featuring aerodynamic curves, racing decals, and a visible driver inside, positioned in the lower centre of the frame - at the start of the clip its front away from the camera. A touring-car style racing saloon with a glossy, dark charcoal body, similar to the camera-owner's car, positioned further back in the middle right of the frame - at the start of the clip its front toward the screen-right. A touring-car style racing saloon with a glossy, dark charcoal body, identical to the other racing sedans, currently behind the camera and not visible in the frame. A large, dark timber gantry structure with a weathered finish, spanning the width of the track, its beams casting long shadows on the grass below. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing upright in the middle ground; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is a realistic adult road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival racing car 1, and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 21 metres along its route; the vehicle moves about 21 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 19 metres along its route. From 00:01.267 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 62 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 55 metres along its route; the vehicle moves about 53 metres along its route; the seated driver of rival racing car 1 moves about 57 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 6> follows <Subject 5> from 00:00.000 to 00:01.100. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 32 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 9 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 3.9 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 4.0 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-082-racing-racing/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-082-racing-racing/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d42f36a48cb6424b23da7dbd204bff192fbd5e11
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-082-racing-racing/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 4.4 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A touring-car style racing saloon with a glossy, dark charcoal body, featuring aerodynamic curves, racing decals, and a visible driver inside, positioned in the lower centre of the frame - at the start of the clip its front away from the camera. A touring-car style racing saloon with a glossy, dark charcoal body, similar to the camera-owner's car, positioned further back in the middle right of the frame - at the start of the clip its front toward the screen-right. A touring-car style racing saloon with a glossy, dark charcoal body, identical to the other racing sedans, currently behind the camera and not visible in the frame. A large, dark timber gantry structure with a weathered finish, spanning the width of the track, its beams casting long shadows on the grass below. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the black solid checkpoint proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver: not in frame at 00:00.000; it enters the frame at about 4.4 s; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 4.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A touring-car style racing saloon with a glossy, dark charcoal body, featuring aerodynamic curves, racing decals, and a visible driver inside, positioned in the lower centre of the frame - at the start of the clip its front away from the camera. A touring-car style racing saloon with a glossy, dark charcoal body, similar to the camera-owner's car, positioned further back in the middle right of the frame - at the start of the clip its front toward the screen-right. A touring-car style racing saloon with a glossy, dark charcoal body, identical to the other racing sedans, currently behind the camera and not visible in the frame. A large, dark timber gantry structure with a weathered finish, spanning the width of the track, its beams casting long shadows on the grass below. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing upright in the middle ground; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is a realistic adult road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival racing car 1, and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 21 metres along its route; the vehicle moves about 21 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 19 metres along its route. From 00:01.267 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 62 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 55 metres along its route; the vehicle moves about 53 metres along its route; the seated driver of rival racing car 1 moves about 57 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 6> follows <Subject 5> from 00:00.000 to 00:01.100. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 32 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 9 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 3.9 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 4.0 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-082-racing-racing/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-082-racing-racing/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d9668bfe8a440521a5c7578412a9a55080131697
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-082-racing-racing/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 4.4 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A touring-car style racing saloon with a glossy, dark charcoal body, featuring aerodynamic curves, racing decals, and a visible driver inside, positioned in the lower centre of the frame - at the start of the clip its front away from the camera. A touring-car style racing saloon with a glossy, dark charcoal body, similar to the camera-owner's car, positioned further back in the middle right of the frame - at the start of the clip its front toward the screen-right. A touring-car style racing saloon with a glossy, dark charcoal body, identical to the other racing sedans, currently behind the camera and not visible in the frame. A large, dark timber gantry structure with a weathered finish, spanning the width of the track, its beams casting long shadows on the grass below. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver: not in frame at 00:00.000; it enters the frame at about 4.4 s; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 4.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A touring-car style racing saloon with a glossy, dark charcoal body, featuring aerodynamic curves, racing decals, and a visible driver inside, positioned in the lower centre of the frame - at the start of the clip its front away from the camera. A touring-car style racing saloon with a glossy, dark charcoal body, similar to the camera-owner's car, positioned further back in the middle right of the frame - at the start of the clip its front toward the screen-right. A touring-car style racing saloon with a glossy, dark charcoal body, identical to the other racing sedans, currently behind the camera and not visible in the frame. A large, dark timber gantry structure with a weathered finish, spanning the width of the track, its beams casting long shadows on the grass below. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing upright in the middle ground; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is a realistic adult road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival racing car 1, and keeps exactly this appearance. Both people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 21 metres along its route; the vehicle moves about 21 metres along its route; the vehicle moves about 18 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 19 metres along its route. From 00:01.267 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 62 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 55 metres along its route; the vehicle moves about 53 metres along its route; the seated driver of rival racing car 1 moves about 57 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 6> follows <Subject 5> from 00:00.000 to 00:01.100. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 32 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 9 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 3.9 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 4.0 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-083-racing-racing/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-083-racing-racing/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..697c62cad6d99860b9f4d7b3056fc0f2ad0f99ea
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-083-racing-racing/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a sleek, dark shade, complete with doors, windows, and wheels, ready for the race, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A touring-car style racing saloon, its body a deep, glossy shade, with a driver inside, poised for the start, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A rally hatchback with mud flaps, its body a rugged, earthy tone, equipped for the rough terrain, with a driver inside, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A large, angular structure made of weathered stone, standing as a silent sentinel at the start of the race, its surface rough and textured with age, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the object proxy tagged #36f7d6; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the human proxy tagged #3ef736; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the human proxy tagged #36aff7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a sleek, dark shade, complete with doors, windows, and wheels, ready for the race, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A touring-car style racing saloon, its body a deep, glossy shade, with a driver inside, poised for the start, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A rally hatchback with mud flaps, its body a rugged, earthy tone, equipped for the rough terrain, with a driver inside, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A large, angular structure made of weathered stone, standing as a silent sentinel at the start of the race, its surface rough and textured with age, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces). <Subject 1>, the protagonist, has the following target appearance: An adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and a practical, weathered appearance, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces); keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is an adult cyclist with short hair under a helmet, wearing an emerald cycling jersey, black cycling shorts, and cleated shoes, focused and ready, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces), and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, exuding authority, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.833, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 2 metres along its route; the vehicle moves about 2 metres along its route; the vehicle moves about one and a half metres along its route; the vehicle moves about a metre along its route; the seated driver of rival racing car 1 moves about one and a half metres along its route; the seated driver of rival racing car 3 moves about a metre along its route. From 00:00.833 to 00:04.867, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 57 metres along its route; the vehicle moves about 57 metres along its route; the vehicle moves about 70 metres along its route; the vehicle moves about 66 metres along its route; the seated driver of rival racing car 1 moves about 70 metres along its route; the seated driver of rival racing car 3 moves about 66 metres along its route. From 00:04.867 to 00:05.000, the protagonist's sedan pulls out alongside a rival sedan and tries to pass while the other rival sedan keeps racing on its own line; the protagonist moves about 2 metres along its route; the vehicle moves about 2 metres along its route; the vehicle moves about 3 metres along its route; the vehicle moves about 3 metres along its route; the seated driver of rival racing car 1 moves about 3 metres along its route; the seated driver of rival racing car 3 moves about 3 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 6> from 00:00.833 to 00:03.900; <Subject 4> overtakes <Subject 6> from 00:02.467 to 00:05.000; <Subject 4> follows <Subject 5> from 00:03.967 to 00:04.867. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 24 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 32 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the protagonist from about 0.1 s to 0.3 s; the same start-and-finish gantry for the whole clip in the foreground covers the protagonist from about 0.5 s to 1.7 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 0.1 s to 0.3 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 0.5 s to 4.3 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 4.3 s to 5.0 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 3.7 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-083-racing-racing/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-083-racing-racing/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4e9582e7ea0e017dadbca60c4d90a02e4ed8b7f9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-083-racing-racing/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a sleek, dark shade, complete with doors, windows, and wheels, ready for the race, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A touring-car style racing saloon, its body a deep, glossy shade, with a driver inside, poised for the start, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A rally hatchback with mud flaps, its body a rugged, earthy tone, equipped for the rough terrain, with a driver inside, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A large, angular structure made of weathered stone, standing as a silent sentinel at the start of the race, its surface rough and textured with age, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the black solid checkpoint proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a sleek, dark shade, complete with doors, windows, and wheels, ready for the race, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A touring-car style racing saloon, its body a deep, glossy shade, with a driver inside, poised for the start, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A rally hatchback with mud flaps, its body a rugged, earthy tone, equipped for the rough terrain, with a driver inside, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A large, angular structure made of weathered stone, standing as a silent sentinel at the start of the race, its surface rough and textured with age, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces). <Subject 1>, the protagonist, has the following target appearance: An adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and a practical, weathered appearance, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces); keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is an adult cyclist with short hair under a helmet, wearing an emerald cycling jersey, black cycling shorts, and cleated shoes, focused and ready, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces), and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, exuding authority, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.833, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 2 metres along its route; the vehicle moves about 2 metres along its route; the vehicle moves about one and a half metres along its route; the vehicle moves about a metre along its route; the seated driver of rival racing car 1 moves about one and a half metres along its route; the seated driver of rival racing car 3 moves about a metre along its route. From 00:00.833 to 00:04.867, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 57 metres along its route; the vehicle moves about 57 metres along its route; the vehicle moves about 70 metres along its route; the vehicle moves about 66 metres along its route; the seated driver of rival racing car 1 moves about 70 metres along its route; the seated driver of rival racing car 3 moves about 66 metres along its route. From 00:04.867 to 00:05.000, the protagonist's sedan pulls out alongside a rival sedan and tries to pass while the other rival sedan keeps racing on its own line; the protagonist moves about 2 metres along its route; the vehicle moves about 2 metres along its route; the vehicle moves about 3 metres along its route; the vehicle moves about 3 metres along its route; the seated driver of rival racing car 1 moves about 3 metres along its route; the seated driver of rival racing car 3 moves about 3 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 6> from 00:00.833 to 00:03.900; <Subject 4> overtakes <Subject 6> from 00:02.467 to 00:05.000; <Subject 4> follows <Subject 5> from 00:03.967 to 00:04.867. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 24 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 32 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the protagonist from about 0.1 s to 0.3 s; the same start-and-finish gantry for the whole clip in the foreground covers the protagonist from about 0.5 s to 1.7 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 0.1 s to 0.3 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 0.5 s to 4.3 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 4.3 s to 5.0 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 3.7 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-083-racing-racing/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-083-racing-racing/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..31c1cc7ebb0f4faffacdf25ddf001ecb5b6eb1ab
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-083-racing-racing/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a sleek, dark shade, complete with doors, windows, and wheels, ready for the race, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A touring-car style racing saloon, its body a deep, glossy shade, with a driver inside, poised for the start, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A rally hatchback with mud flaps, its body a rugged, earthy tone, equipped for the rough terrain, with a driver inside, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A large, angular structure made of weathered stone, standing as a silent sentinel at the start of the race, its surface rough and textured with age, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Sand dunes, angular desert rocks, cactus plants and roofless stone ruins. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, painted in a sleek, dark shade, complete with doors, windows, and wheels, ready for the race, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A touring-car style racing saloon, its body a deep, glossy shade, with a driver inside, poised for the start, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A rally hatchback with mud flaps, its body a rugged, earthy tone, equipped for the rough terrain, with a driver inside, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A large, angular structure made of weathered stone, standing as a silent sentinel at the start of the race, its surface rough and textured with age, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces). <Subject 1>, the protagonist, has the following target appearance: An adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, and a practical, weathered appearance, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces); keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is an adult cyclist with short hair under a helmet, wearing an emerald cycling jersey, black cycling shorts, and cleated shoes, focused and ready, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces), and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, exuding authority, drawn as an anime character in the same watercolour-background anime manner (character design: realistically proportioned bodies with simplified anime faces), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.833, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 2 metres along its route; the vehicle moves about 2 metres along its route; the vehicle moves about one and a half metres along its route; the vehicle moves about a metre along its route; the seated driver of rival racing car 1 moves about one and a half metres along its route; the seated driver of rival racing car 3 moves about a metre along its route. From 00:00.833 to 00:04.867, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 57 metres along its route; the vehicle moves about 57 metres along its route; the vehicle moves about 70 metres along its route; the vehicle moves about 66 metres along its route; the seated driver of rival racing car 1 moves about 70 metres along its route; the seated driver of rival racing car 3 moves about 66 metres along its route. From 00:04.867 to 00:05.000, the protagonist's sedan pulls out alongside a rival sedan and tries to pass while the other rival sedan keeps racing on its own line; the protagonist moves about 2 metres along its route; the vehicle moves about 2 metres along its route; the vehicle moves about 3 metres along its route; the vehicle moves about 3 metres along its route; the seated driver of rival racing car 1 moves about 3 metres along its route; the seated driver of rival racing car 3 moves about 3 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 6> from 00:00.833 to 00:03.900; <Subject 4> overtakes <Subject 6> from 00:02.467 to 00:05.000; <Subject 4> follows <Subject 5> from 00:03.967 to 00:04.867. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 24 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 32 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the protagonist from about 0.1 s to 0.3 s; the same start-and-finish gantry for the whole clip in the foreground covers the protagonist from about 0.5 s to 1.7 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 0.1 s to 0.3 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 0.5 s to 4.3 s; the protagonist's racing sedan with a seated driver in the foreground covers the protagonist from about 4.3 s to 5.0 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 3.7 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — watercolour-background anime: softly shaded characters over loose watercolour-washed backgrounds, light contour lines and airy pastel colours; a moody deep-shadow palette with luminous highlights, flat overcast daylight, gentle wind moving loose fabric and foliage; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-087-racing-racing/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-087-racing-racing/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8c3324180726e7c4ec55ed4c805914872f2482f6
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-087-racing-racing/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 4.2 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, featuring a glossy charcoal body, aerodynamic doors, tinted windows, and racing wheels, positioned in the immediate foreground - at the start of the clip its front away from the camera. A touring-car style racing saloon with a matte slate body, complete with doors, windows, and wheels, driving away in the middle distance. A low racing sedan with a rear wing, featuring a dark grey body and racing livery, visible further down the track. A massive, heavy-duty steel gantry structure with riveted beams and a weathered grey finish, spanning the width of the track overhead. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1: not in frame at 00:00.000; it enters the frame at about 4.2 s; in <Video 1> the human proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the human proxy tagged #36aff7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 4.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, featuring a glossy charcoal body, aerodynamic doors, tinted windows, and racing wheels, positioned in the immediate foreground - at the start of the clip its front away from the camera. A touring-car style racing saloon with a matte slate body, complete with doors, windows, and wheels, driving away in the middle distance. A low racing sedan with a rear wing, featuring a dark grey body and racing livery, visible further down the track. A massive, heavy-duty steel gantry structure with riveted beams and a weathered grey finish, spanning the width of the track overhead. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing upright in the center of the frame; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is not in frame at 00:00.000 and enters the frame at about 4.2 s; from then on it is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes and charcoal uniform trousers, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.400, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 31 metres along its route; the vehicle moves about 31 metres along its route; the vehicle moves about 26 metres along its route; the vehicle moves about 46 metres along its route; the seated driver of rival racing car 1 moves about 26 metres along its route; the seated driver of rival racing car 3 moves about 46 metres along its route. From 00:01.400 to 00:04.467, the protagonist's sedan pulls out and passes a rival sedan while the other rival sedan keeps racing on its own line; the protagonist moves about 86 metres along its route; the vehicle moves about 86 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 110 metres along its route; the seated driver of rival racing car 1 moves about 61 metres along its route; the seated driver of rival racing car 3 moves about 110 metres along its route. From 00:04.467 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 10 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 10 metres along its route; the seated driver of rival racing car 3 moves about 18 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 5> from 00:00.000 to 00:01.400; <Subject 4> overtakes <Subject 5> from 00:00.000 to 00:04.467. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 17 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 106 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.3 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-087-racing-racing/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-087-racing-racing/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..128dda02d4c7e76a0b67456f0f25352c72d76b2e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-087-racing-racing/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 4.2 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, featuring a glossy charcoal body, aerodynamic doors, tinted windows, and racing wheels, positioned in the immediate foreground - at the start of the clip its front away from the camera. A touring-car style racing saloon with a matte slate body, complete with doors, windows, and wheels, driving away in the middle distance. A low racing sedan with a rear wing, featuring a dark grey body and racing livery, visible further down the track. A massive, heavy-duty steel gantry structure with riveted beams and a weathered grey finish, spanning the width of the track overhead. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the black solid checkpoint proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1: not in frame at 00:00.000; it enters the frame at about 4.2 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 4.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, featuring a glossy charcoal body, aerodynamic doors, tinted windows, and racing wheels, positioned in the immediate foreground - at the start of the clip its front away from the camera. A touring-car style racing saloon with a matte slate body, complete with doors, windows, and wheels, driving away in the middle distance. A low racing sedan with a rear wing, featuring a dark grey body and racing livery, visible further down the track. A massive, heavy-duty steel gantry structure with riveted beams and a weathered grey finish, spanning the width of the track overhead. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing upright in the center of the frame; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is not in frame at 00:00.000 and enters the frame at about 4.2 s; from then on it is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes and charcoal uniform trousers, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.400, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 31 metres along its route; the vehicle moves about 31 metres along its route; the vehicle moves about 26 metres along its route; the vehicle moves about 46 metres along its route; the seated driver of rival racing car 1 moves about 26 metres along its route; the seated driver of rival racing car 3 moves about 46 metres along its route. From 00:01.400 to 00:04.467, the protagonist's sedan pulls out and passes a rival sedan while the other rival sedan keeps racing on its own line; the protagonist moves about 86 metres along its route; the vehicle moves about 86 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 110 metres along its route; the seated driver of rival racing car 1 moves about 61 metres along its route; the seated driver of rival racing car 3 moves about 110 metres along its route. From 00:04.467 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 10 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 10 metres along its route; the seated driver of rival racing car 3 moves about 18 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 5> from 00:00.000 to 00:01.400; <Subject 4> overtakes <Subject 5> from 00:00.000 to 00:04.467. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 17 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 106 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.3 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-087-racing-racing/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-087-racing-racing/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e880449b92434ce0056a0ce21bff09fdf6d072d8
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-087-racing-racing/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 4.2 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, featuring a glossy charcoal body, aerodynamic doors, tinted windows, and racing wheels, positioned in the immediate foreground - at the start of the clip its front away from the camera. A touring-car style racing saloon with a matte slate body, complete with doors, windows, and wheels, driving away in the middle distance. A low racing sedan with a rear wing, featuring a dark grey body and racing livery, visible further down the track. A massive, heavy-duty steel gantry structure with riveted beams and a weathered grey finish, spanning the width of the track overhead. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1: not in frame at 00:00.000; it enters the frame at about 4.2 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 4.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, featuring a glossy charcoal body, aerodynamic doors, tinted windows, and racing wheels, positioned in the immediate foreground - at the start of the clip its front away from the camera. A touring-car style racing saloon with a matte slate body, complete with doors, windows, and wheels, driving away in the middle distance. A low racing sedan with a rear wing, featuring a dark grey body and racing livery, visible further down the track. A massive, heavy-duty steel gantry structure with riveted beams and a weathered grey finish, spanning the width of the track overhead. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing upright in the center of the frame; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is not in frame at 00:00.000 and enters the frame at about 4.2 s; from then on it is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes and charcoal uniform trousers, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.400, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 31 metres along its route; the vehicle moves about 31 metres along its route; the vehicle moves about 26 metres along its route; the vehicle moves about 46 metres along its route; the seated driver of rival racing car 1 moves about 26 metres along its route; the seated driver of rival racing car 3 moves about 46 metres along its route. From 00:01.400 to 00:04.467, the protagonist's sedan pulls out and passes a rival sedan while the other rival sedan keeps racing on its own line; the protagonist moves about 86 metres along its route; the vehicle moves about 86 metres along its route; the vehicle moves about 61 metres along its route; the vehicle moves about 110 metres along its route; the seated driver of rival racing car 1 moves about 61 metres along its route; the seated driver of rival racing car 3 moves about 110 metres along its route. From 00:04.467 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 19 metres along its route; the vehicle moves about 19 metres along its route; the vehicle moves about 10 metres along its route; the vehicle moves about 18 metres along its route; the seated driver of rival racing car 1 moves about 10 metres along its route; the seated driver of rival racing car 3 moves about 18 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; the seated driver of rival racing car 3 occupies <Subject 6> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 5> from 00:00.000 to 00:01.400; <Subject 4> overtakes <Subject 5> from 00:00.000 to 00:04.467. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 17 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 106 metres from the protagonist. The same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 0.4 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.0 s to 4.3 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-088-racing-racing/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-088-racing-racing/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fbc677635683be7c81f6833b5d3be27f0e000235
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-088-racing-racing/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 4.7 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy metallic silver with teal racing stripes, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit. A towering steel gantry structure with a dark metallic finish, featuring glowing amber lights and sharp, hard-edged mechanical detailing. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the object proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver: not in frame at 00:00.000; it enters the frame at about 4.7 s; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the human proxy tagged #3ef736 that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the human proxy tagged #36aff7 that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 4.7 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy metallic silver with teal racing stripes, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit. A towering steel gantry structure with a dark metallic finish, featuring glowing amber lights and sharp, hard-edged mechanical detailing. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is a road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.300, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 22 metres along its route; the vehicle moves about 22 metres along its route; the vehicle moves about 20 metres along its route; the vehicle moves about 19 metres along its route; the seated driver of rival racing car 1 moves about 20 metres along its route; the seated driver of rival racing car 3 moves about 22 metres along its route. From 00:01.300 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 63 metres along its route; the vehicle moves about 62 metres along its route; the vehicle moves about 57 metres along its route; the vehicle moves about 54 metres along its route; the seated driver of rival racing car 1 moves about 58 metres along its route; the seated driver of rival racing car 3 moves about 65 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 6> follows <Subject 5> from 00:00.000 to 00:01.400. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 33 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 10 metres from the protagonist. The short centre stripe and paired edge kerbs in the foreground covers the seated driver of rival racing car 1 from about 0.5 s to 0.7 s; the same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.3 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.3 s to 2.3 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 2.5 s to 4.2 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 4.4 s to 5.0 s; the tree with a distinct trunk and crown in the foreground covers the seated driver of rival racing car 3 from about 3.3 s to 4.0 s; the same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 3 from about 0.1 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 3 from about 0.1 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-088-racing-racing/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-088-racing-racing/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..138a5288211b15bdc2554b36f7164edf3601f6af
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-088-racing-racing/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 4.7 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy metallic silver with teal racing stripes, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit. A towering steel gantry structure with a dark metallic finish, featuring glowing amber lights and sharp, hard-edged mechanical detailing. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the black solid checkpoint proxy shape; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver: not in frame at 00:00.000; it enters the frame at about 4.7 s; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 4.7 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy metallic silver with teal racing stripes, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit. A towering steel gantry structure with a dark metallic finish, featuring glowing amber lights and sharp, hard-edged mechanical detailing. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is a road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.300, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 22 metres along its route; the vehicle moves about 22 metres along its route; the vehicle moves about 20 metres along its route; the vehicle moves about 19 metres along its route; the seated driver of rival racing car 1 moves about 20 metres along its route; the seated driver of rival racing car 3 moves about 22 metres along its route. From 00:01.300 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 63 metres along its route; the vehicle moves about 62 metres along its route; the vehicle moves about 57 metres along its route; the vehicle moves about 54 metres along its route; the seated driver of rival racing car 1 moves about 58 metres along its route; the seated driver of rival racing car 3 moves about 65 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 6> follows <Subject 5> from 00:00.000 to 00:01.400. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 33 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 10 metres from the protagonist. The short centre stripe and paired edge kerbs in the foreground covers the seated driver of rival racing car 1 from about 0.5 s to 0.7 s; the same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.3 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.3 s to 2.3 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 2.5 s to 4.2 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 4.4 s to 5.0 s; the tree with a distinct trunk and crown in the foreground covers the seated driver of rival racing car 3 from about 3.3 s to 4.0 s; the same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 3 from about 0.1 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 3 from about 0.1 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-088-racing-racing/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-088-racing-racing/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bcc68e0d81f675df1d12fd26e427ba6c3513f4ae
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-088-racing-racing/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 7> (the seated driver of rival racing car 1), <Subject 8> (the seated driver of rival racing car 3) and 4 object proxies standing for <Subject 3> (the same start-and-finish gantry for the whole clip), <Subject 4> (the protagonist's racing sedan with a seated driver), <Subject 5> (the rival sedan with a seated driver), <Subject 6> (the rival sedan with a seated driver); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 4.7 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy metallic silver with teal racing stripes, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit. A towering steel gantry structure with a dark metallic finish, featuring glowing amber lights and sharp, hard-edged mechanical detailing. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the same start-and-finish gantry for the whole clip; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the protagonist's racing sedan with a seated driver; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the rival sedan with a seated driver; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the rival sedan with a seated driver: not in frame at 00:00.000; it enters the frame at about 4.7 s; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the seated driver of rival racing car 1; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the seated driver of rival racing car 3; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> racing a sedan along the track against two rival sedans, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 4.7 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A low racing sedan with a rear wing, finished in glossy metallic silver with teal racing stripes, featuring a complete body with doors, windows, and aerodynamic wheels - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit - at the start of the clip its front away from the camera. A low racing sedan with a rear wing, finished in matte charcoal grey with crimson accents, appearing as a complete vehicle with a visible cockpit. A towering steel gantry structure with a dark metallic finish, featuring glowing amber lights and sharp, hard-edged mechanical detailing. <Subject 1>, the protagonist, has the following target appearance: A mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the same start-and-finish gantry for the whole clip, realized with the materials above. <Subject 4> is the protagonist's racing sedan with a seated driver, realized with the materials above. <Subject 5> is the rival sedan with a seated driver, realized with the materials above. <Subject 6> is the rival sedan with a seated driver, realized with the materials above. <Subject 7> (the seated driver of rival racing car 1) is a road cyclist with short hair under a helmet, wearing an emerald cycling jersey and black cycling shorts, seated inside the rival vehicle, and keeps exactly this appearance. <Subject 8> (the seated driver of rival racing car 3) is one of the people described above, with one consistent appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.300, the protagonist's sedan follows the rival sedan ahead along the track, holding its line, while the other rival sedan keeps racing on its own line; the protagonist moves about 22 metres along its route; the vehicle moves about 22 metres along its route; the vehicle moves about 20 metres along its route; the vehicle moves about 19 metres along its route; the seated driver of rival racing car 1 moves about 20 metres along its route; the seated driver of rival racing car 3 moves about 22 metres along its route. From 00:01.300 to 00:05.000, the protagonist's sedan cruises along the track at steady speed while the two rival sedans keep racing on their own lines; the protagonist moves about 63 metres along its route; the vehicle moves about 62 metres along its route; the vehicle moves about 57 metres along its route; the vehicle moves about 54 metres along its route; the seated driver of rival racing car 1 moves about 58 metres along its route; the seated driver of rival racing car 3 moves about 65 metres along its route. Recorded relations: the protagonist occupies the protagonist's racing sedan with a seated driver from 00:00.000 to 00:05.000; the seated driver of rival racing car 1 occupies <Subject 5> from 00:00.000 to 00:05.000; <Subject 6> follows <Subject 5> from 00:00.000 to 00:01.400. The protagonist's racing sedan with a seated driver is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 33 metres from the protagonist. <Subject 6> is one continuous object throughout and ends the clip about 10 metres from the protagonist. The short centre stripe and paired edge kerbs in the foreground covers the seated driver of rival racing car 1 from about 0.5 s to 0.7 s; the same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 1 from about 0.3 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 0.3 s to 2.3 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 2.5 s to 4.2 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 1 from about 4.4 s to 5.0 s; the tree with a distinct trunk and crown in the foreground covers the seated driver of rival racing car 3 from about 3.3 s to 4.0 s; the same start-and-finish gantry for the whole clip in the foreground covers the seated driver of rival racing car 3 from about 0.1 s to 0.5 s; the rival sedan with a seated driver in the foreground covers the seated driver of rival racing car 3 from about 0.1 s to 5.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..eaef4c417db06ee6b725ba1b7792f23c694003e6
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the sliding door), <Subject 4> (the switch); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 2.3 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.8 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular sliding door made of smooth, matte sandstone with a simple, dark square handle, set flush into the rock wall, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). A small, square control panel with a dark, recessed button, mounted on the wall surface, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the sliding door; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the switch; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1: not in frame at 00:00.000; it enters the frame at about 2.3 s; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the human proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #6636f7 that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> pressing a control to open a sliding gate and passing through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 2.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 6> (enters [Shot 1] at about 2.8 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular sliding door made of smooth, matte sandstone with a simple, dark square handle, set flush into the rock wall, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). A small, square control panel with a dark, recessed button, mounted on the wall surface, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). <Subject 1>, the protagonist, has the following target appearance: A lanky, noodle-limbed adult explorer with a tiny head, featuring long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the sliding door, realized with the materials above. <Subject 4> is the switch, realized with the materials above. <Subject 5> (independently moving adult participant 1) is not in frame at 00:00.000 and enters the frame at about 2.3 s; from then on it is a lanky, noodle-limbed adult mechanic with a tiny head, featuring auburn hair tied back, wearing navy work overalls and steel-toe boots, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is not in frame at 00:00.000 and enters the frame at about 2.8 s; from then on it is a lanky, noodle-limbed adult tourist with a tiny head, featuring curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a lanky, noodle-limbed adult nurse with a tiny head, featuring dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the sliding gate and its control; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:00.967 to 00:02.067, the protagonist presses the control and the gate slides open; the protagonist moves about a metre along its route, lowers the hands to the sides; the door moves about 3 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open gate and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the protagonist opens the sliding door from 00:00.000 to 00:02.067; the protagonist passed through the sliding door from 00:02.067 to 00:05.000. The sliding door is one continuous object throughout and comes to rest about 5 metres from the protagonist. The sliding door in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.2 s; the door-frame-1 in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.3 s; the door-frame-1 in the foreground covers independently moving adult participant 1 from about 4.7 s to 5.0 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 0.0 s to 2.9 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 4.3 s to 5.0 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 2 from about 1.3 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3d7bcb95da5e0e69a0db0da7755e97622e999b38
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the sliding door), <Subject 4> (the switch); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 2.3 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.8 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular sliding door made of smooth, matte sandstone with a simple, dark square handle, set flush into the rock wall, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). A small, square control panel with a dark, recessed button, mounted on the wall surface, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the sliding door; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the switch; in <Video 1> the black solid control-panel proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1: not in frame at 00:00.000; it enters the frame at about 2.3 s; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> pressing a control to open a sliding gate and passing through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 2.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 6> (enters [Shot 1] at about 2.8 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular sliding door made of smooth, matte sandstone with a simple, dark square handle, set flush into the rock wall, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). A small, square control panel with a dark, recessed button, mounted on the wall surface, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). <Subject 1>, the protagonist, has the following target appearance: A lanky, noodle-limbed adult explorer with a tiny head, featuring long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the sliding door, realized with the materials above. <Subject 4> is the switch, realized with the materials above. <Subject 5> (independently moving adult participant 1) is not in frame at 00:00.000 and enters the frame at about 2.3 s; from then on it is a lanky, noodle-limbed adult mechanic with a tiny head, featuring auburn hair tied back, wearing navy work overalls and steel-toe boots, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is not in frame at 00:00.000 and enters the frame at about 2.8 s; from then on it is a lanky, noodle-limbed adult tourist with a tiny head, featuring curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a lanky, noodle-limbed adult nurse with a tiny head, featuring dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the sliding gate and its control; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:00.967 to 00:02.067, the protagonist presses the control and the gate slides open; the protagonist moves about a metre along its route, lowers the hands to the sides; the door moves about 3 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open gate and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the protagonist opens the sliding door from 00:00.000 to 00:02.067; the protagonist passed through the sliding door from 00:02.067 to 00:05.000. The sliding door is one continuous object throughout and comes to rest about 5 metres from the protagonist. The sliding door in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.2 s; the door-frame-1 in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.3 s; the door-frame-1 in the foreground covers independently moving adult participant 1 from about 4.7 s to 5.0 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 0.0 s to 2.9 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 4.3 s to 5.0 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 2 from about 1.3 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..85677623a659c1f628b2442dfcf306c42e30de9f
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-089-fantasy-gate/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the sliding door), <Subject 4> (the switch); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 2.3 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.8 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular sliding door made of smooth, matte sandstone with a simple, dark square handle, set flush into the rock wall, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). A small, square control panel with a dark, recessed button, mounted on the wall surface, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the sliding door; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the switch; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1: not in frame at 00:00.000; it enters the frame at about 2.3 s; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2: not in frame at 00:00.000; it enters the frame at about 2.8 s; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> pressing a control to open a sliding gate and passing through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 2.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (enters [Shot 1] at about 2.8 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular sliding door made of smooth, matte sandstone with a simple, dark square handle, set flush into the rock wall, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). A small, square control panel with a dark, recessed button, mounted on the wall surface, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings). <Subject 1>, the protagonist, has the following target appearance: A lanky, noodle-limbed adult explorer with a tiny head, featuring long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the sliding door, realized with the materials above. <Subject 4> is the switch, realized with the materials above. <Subject 5> (independently moving adult participant 1) is not in frame at 00:00.000 and enters the frame at about 2.3 s; from then on it is a lanky, noodle-limbed adult mechanic with a tiny head, featuring auburn hair tied back, wearing navy work overalls and steel-toe boots, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is not in frame at 00:00.000 and enters the frame at about 2.8 s; from then on it is a lanky, noodle-limbed adult tourist with a tiny head, featuring curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a lanky, noodle-limbed adult nurse with a tiny head, featuring dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, seen only from behind, drawn in the same coloured pencil sketch manner (character design: figures as clean contour drawings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the sliding gate and its control; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:00.967 to 00:02.067, the protagonist presses the control and the gate slides open; the protagonist moves about a metre along its route, lowers the hands to the sides; the door moves about 3 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open gate and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the protagonist opens the sliding door from 00:00.000 to 00:02.067; the protagonist passed through the sliding door from 00:02.067 to 00:05.000. The sliding door is one continuous object throughout and comes to rest about 5 metres from the protagonist. The sliding door in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.2 s; the door-frame-1 in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.3 s; the door-frame-1 in the foreground covers independently moving adult participant 1 from about 4.7 s to 5.0 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 0.0 s to 2.9 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 4.3 s to 5.0 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 2 from about 1.3 s to 1.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, coloured pencil sketch look: layered pencil hatching in muted colours over a white paper ground, visible paper grain, hand-drawn contour lines, drawn over the full 3D world so every recorded shape stays readable; a low-saturation earthy palette, flat overcast daylight, crisp cold air with a hint of frost, background fully hatched in lighter tone; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c66b043972dc3f7b39ad5fa1004c3499781cc0c3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person receiving the object), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.7 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cargo crate with a wooden finish and metal reinforcements, held by the protagonist, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (enters [Shot 1] at about 0.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cargo crate with a wooden finish and metal reinforcements, held by the protagonist, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines). <Subject 1>, the protagonist, has the following target appearance: An adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the middle centre of the frame, seen only from behind, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, positioned slightly to the right of the player, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the far left of the frame, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, positioned further back and to the left, seen only from behind, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is not in frame at 00:00.000 and enters the frame at about 0.7 s; from then on it is an adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, not yet in frame but will enter later, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The irregular cluster of solid crystals in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.3 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.0 s to 0.8 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 3 from about 0.0 s to 0.8 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 3 from about 4.3 s to 4.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a05a1825f3450a416d3c67db7af08d28c9c7b0b3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person receiving the object), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.7 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cargo crate with a wooden finish and metal reinforcements, held by the protagonist, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (enters [Shot 1] at about 0.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cargo crate with a wooden finish and metal reinforcements, held by the protagonist, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines). <Subject 1>, the protagonist, has the following target appearance: An adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the middle centre of the frame, seen only from behind, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, positioned slightly to the right of the player, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the far left of the frame, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, positioned further back and to the left, seen only from behind, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is not in frame at 00:00.000 and enters the frame at about 0.7 s; from then on it is an adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, not yet in frame but will enter later, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The irregular cluster of solid crystals in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.3 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.0 s to 0.8 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 3 from about 0.0 s to 0.8 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 3 from about 4.3 s to 4.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6a41d216992546888e9b092a9bbba1f7dbf42233
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-090-fantasy-handoff/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person receiving the object), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.7 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cargo crate with a wooden finish and metal reinforcements, held by the protagonist, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3: not in frame at 00:00.000; it enters the frame at about 0.7 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (enters [Shot 1] at about 0.7 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky cave with irregular walls, sloping ground, pillars, low ceilings and an entrance platform. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cargo crate with a wooden finish and metal reinforcements, held by the protagonist, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines). <Subject 1>, the protagonist, has the following target appearance: An adult explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing in the middle centre of the frame, seen only from behind, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, positioned slightly to the right of the player, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the far left of the frame, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, positioned further back and to the left, seen only from behind, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is not in frame at 00:00.000 and enters the frame at about 0.7 s; from then on it is an adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, not yet in frame but will enter later, drawn as an anime character in the same gouache-textured anime manner (character design: compact stocky figures with thick outlines), and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The irregular cluster of solid crystals in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.3 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.0 s to 0.8 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 3 from about 0.0 s to 0.8 s; the irregular cluster of solid crystals in the foreground covers independently moving adult participant 3 from about 4.3 s to 4.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — gouache-textured anime look: opaque matte paint textures on backgrounds, characters with soft two-tone shading, warm grainy paper feel; a cool blue-grey palette with muted warm accents, golden-hour side light with long soft shadows, light drifting haze, clean uniform outlines; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b8caa18e3e9cbbce5ff48a56dd9a4d28a620ddee
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person catching the ball), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the throwable ball); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A spherical object made of stitched crimson felt with a visible seam running around its circumference, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the throwable ball; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the person catching the ball; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #6636f7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> throwing a ball through the air to a second person who catches it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A spherical object made of stitched crimson felt with a visible seam running around its circumference, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs). <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the throwable ball, realized with the materials above. <Subject 4> (the person catching the ball) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, seen only from behind, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist winds up and throws the ball toward the second person; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:01.267 to 00:02.367, the ball flies through the air toward the second person; the protagonist lowers the hands; the ball moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.367 to 00:05.000, the second person catches the ball and holds it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the person catching the ball holds the throwable ball from 00:02.367 to 00:05.000. The throwable ball is one continuous object throughout and comes to rest about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 0.4 s to 0.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5f96b0127f4eee6c29bccee6c4c0af1a80f7a0e2
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person catching the ball), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the throwable ball); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A spherical object made of stitched crimson felt with a visible seam running around its circumference, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the throwable ball; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the person catching the ball; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> throwing a ball through the air to a second person who catches it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A spherical object made of stitched crimson felt with a visible seam running around its circumference, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs). <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the throwable ball, realized with the materials above. <Subject 4> (the person catching the ball) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, seen only from behind, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist winds up and throws the ball toward the second person; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:01.267 to 00:02.367, the ball flies through the air toward the second person; the protagonist lowers the hands; the ball moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.367 to 00:05.000, the second person catches the ball and holds it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the person catching the ball holds the throwable ball from 00:02.367 to 00:05.000. The throwable ball is one continuous object throughout and comes to rest about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 0.4 s to 0.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9a7b244adfd0a7b06dc98f327502e8fbf8d09970
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-092-fantasy-throw/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (the person catching the ball), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the throwable ball); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A spherical object made of stitched crimson felt with a visible seam running around its circumference, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the throwable ball; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the person catching the ball; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> throwing a ball through the air to a second person who catches it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A spherical object made of stitched crimson felt with a visible seam running around its circumference, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs). <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the throwable ball, realized with the materials above. <Subject 4> (the person catching the ball) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, seen only from behind, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, drawn in the same flat vector cartoon manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist winds up and throws the ball toward the second person; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:01.267 to 00:02.367, the ball flies through the air toward the second person; the protagonist lowers the hands; the ball moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. From 00:02.367 to 00:05.000, the second person catches the ball and holds it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the person catching the ball holds the throwable ball from 00:02.367 to 00:05.000. The throwable ball is one continuous object throughout and comes to rest about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 0.4 s to 0.6 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, flat vector cartoon look: crisp geometric shapes, flat colour fills with no gradients, bold simplified figures and minimal outlines, drawn over the full 3D world so every recorded shape stays readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, clear dry air, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..61f3802a15e593a854cf054c273432919899bd69
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the grazing or ridden unicorn with a horse body and a single horn), <Subject 4> (the grazing or ridden unicorn with a horse body and a single horn), <Subject 5> (the grazing or ridden unicorn with a horse body and a single horn); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A white unicorn-like horse with a single spiral horn, possessing a pearlescent coat and a leather saddle, grazing in the lower right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A white unicorn-like horse with a single spiral horn, with a pearlescent coat and a leather saddle, standing further back in the middle right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A majestic white unicorn-like horse with a single spiral horn, featuring a smooth, pearlescent coat and a leather saddle, standing prominently in the lower centre, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the object proxy tagged #6636f7 that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the object proxy tagged #3ef736 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse while horses and pedestrians move nearby, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A white unicorn-like horse with a single spiral horn, possessing a pearlescent coat and a leather saddle, grazing in the lower right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A white unicorn-like horse with a single spiral horn, with a pearlescent coat and a leather saddle, standing further back in the middle right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A majestic white unicorn-like horse with a single spiral horn, featuring a smooth, pearlescent coat and a leather saddle, standing prominently in the lower centre, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots, standing in the foreground, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 4> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 5> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult farmer with grey hair under a flat cap, wearing a faded denim shirt, canvas trousers, and muddy rubber boots, standing in the middle left, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an adult student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing in the middle centre, seen only from behind, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is an adult field explorer with long braided dark hair, wearing a field shirt with rolled sleeves, cargo trousers, and leather boots, standing in the distance, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward while the other horses and the pedestrians move nearby; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward among the moving horses and pedestrians; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 4 metres from the protagonist. The paddock fence in the foreground covers independently moving adult participant 1 from about 3.0 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f8e783d91c0f635ab46db5911b579217cb6e6314
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the grazing or ridden unicorn with a horse body and a single horn), <Subject 4> (the grazing or ridden unicorn with a horse body and a single horn), <Subject 5> (the grazing or ridden unicorn with a horse body and a single horn); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A white unicorn-like horse with a single spiral horn, possessing a pearlescent coat and a leather saddle, grazing in the lower right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A white unicorn-like horse with a single spiral horn, with a pearlescent coat and a leather saddle, standing further back in the middle right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A majestic white unicorn-like horse with a single spiral horn, featuring a smooth, pearlescent coat and a leather saddle, standing prominently in the lower centre, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse while horses and pedestrians move nearby, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A white unicorn-like horse with a single spiral horn, possessing a pearlescent coat and a leather saddle, grazing in the lower right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A white unicorn-like horse with a single spiral horn, with a pearlescent coat and a leather saddle, standing further back in the middle right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A majestic white unicorn-like horse with a single spiral horn, featuring a smooth, pearlescent coat and a leather saddle, standing prominently in the lower centre, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots, standing in the foreground, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 4> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 5> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult farmer with grey hair under a flat cap, wearing a faded denim shirt, canvas trousers, and muddy rubber boots, standing in the middle left, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an adult student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing in the middle centre, seen only from behind, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is an adult field explorer with long braided dark hair, wearing a field shirt with rolled sleeves, cargo trousers, and leather boots, standing in the distance, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward while the other horses and the pedestrians move nearby; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward among the moving horses and pedestrians; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 4 metres from the protagonist. The paddock fence in the foreground covers independently moving adult participant 1 from about 3.0 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..34da69673192ffeb0d5b999943579fb99569711a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-097-fantasy-herd/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 3 animal proxies (horse) standing for <Subject 3> (the grazing or ridden unicorn with a horse body and a single horn), <Subject 4> (the grazing or ridden unicorn with a horse body and a single horn), <Subject 5> (the grazing or ridden unicorn with a horse body and a single horn); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A white unicorn-like horse with a single spiral horn, possessing a pearlescent coat and a leather saddle, grazing in the lower right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A white unicorn-like horse with a single spiral horn, with a pearlescent coat and a leather saddle, standing further back in the middle right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A majestic white unicorn-like horse with a single spiral horn, featuring a smooth, pearlescent coat and a leather saddle, standing prominently in the lower centre, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the grazing or ridden unicorn with a horse body and a single horn; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> riding a horse while horses and pedestrians move nearby, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A white unicorn-like horse with a single spiral horn, possessing a pearlescent coat and a leather saddle, grazing in the lower right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A white unicorn-like horse with a single spiral horn, with a pearlescent coat and a leather saddle, standing further back in the middle right, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. A majestic white unicorn-like horse with a single spiral horn, featuring a smooth, pearlescent coat and a leather saddle, standing prominently in the lower centre, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult park ranger with short grey hair under a wide-brim hat, wearing a ranger shirt, trousers, and boots, standing in the foreground, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip its front away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 4> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 5> is the grazing or ridden unicorn with a horse body and a single horn, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult farmer with grey hair under a flat cap, wearing a faded denim shirt, canvas trousers, and muddy rubber boots, standing in the middle left, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an adult student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing in the middle centre, seen only from behind, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is an adult field explorer with long braided dark hair, wearing a field shirt with rolled sleeves, cargo trousers, and leather boots, standing in the distance, painted in the same long-scroll panorama ink wash manner (character design: small figures dwarfed by the surroundings) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:02.500, the protagonist rides the horse forward while the other horses and the pedestrians move nearby; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. From 00:02.500 to 00:05.000, the protagonist keeps riding forward among the moving horses and pedestrians; the protagonist moves about 4 metres along its route; the horse moves about 4 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the horse moves about one and a half metres along its route; the horse moves about one and a half metres along its route. Recorded relations: the protagonist rides <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip a short distance from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 4 metres from the protagonist. The paddock fence in the foreground covers independently moving adult participant 1 from about 3.0 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, long-scroll ink look: stacked receding planes of hills, water and mist, tiny figures scaled against a vast landscape, graded grey washes with a few dark accents, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a graded grey tonal palette, harsh midday sun with crisp shadows, thin fog hugging the ground, long sweeping wet brush strokes, mist separating the depth planes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..766ab7b377a1cf2349654e49d5932b9160278493
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the sliding door), <Subject 4> (the switch); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A massive, flat sliding door made of smooth, unadorned sandstone, set flush within the dark, recessed opening of the ruin. A small, rectangular control panel with a metallic finish, mounted on the wall to the right of the sliding door. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the sliding door; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the switch; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> pressing a control to open a sliding gate and passing through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A massive, flat sliding door made of smooth, unadorned sandstone, set flush within the dark, recessed opening of the ruin. A small, rectangular control panel with a metallic finish, mounted on the wall to the right of the sliding door. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing centrally in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the sliding door, realized with the materials above. <Subject 4> is the switch, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing to the left of the central figure - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing to the far right of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back on the far left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the sliding gate and its control; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:00.967 to 00:02.067, the protagonist presses the control and the gate slides open; the protagonist moves about a metre along its route, lowers the hands to the sides; the door moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open gate and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist opens the sliding door from 00:00.000 to 00:02.067; the protagonist passed through the sliding door from 00:02.067 to 00:05.000. The sliding door is one continuous object throughout and comes to rest about 5 metres from the protagonist. The irregular faceted rock in the foreground covers independently moving adult participant 2 from about 2.8 s to 3.8 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 2.8 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.4 s to 1.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..050028420b5cd4a3ac191c251c9b1aa6ce5f2f3e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the sliding door), <Subject 4> (the switch); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A massive, flat sliding door made of smooth, unadorned sandstone, set flush within the dark, recessed opening of the ruin. A small, rectangular control panel with a metallic finish, mounted on the wall to the right of the sliding door. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the sliding door; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the switch; in <Video 1> the black solid control-panel proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> pressing a control to open a sliding gate and passing through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A massive, flat sliding door made of smooth, unadorned sandstone, set flush within the dark, recessed opening of the ruin. A small, rectangular control panel with a metallic finish, mounted on the wall to the right of the sliding door. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing centrally in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the sliding door, realized with the materials above. <Subject 4> is the switch, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing to the left of the central figure - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing to the far right of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back on the far left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the sliding gate and its control; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:00.967 to 00:02.067, the protagonist presses the control and the gate slides open; the protagonist moves about a metre along its route, lowers the hands to the sides; the door moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open gate and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist opens the sliding door from 00:00.000 to 00:02.067; the protagonist passed through the sliding door from 00:02.067 to 00:05.000. The sliding door is one continuous object throughout and comes to rest about 5 metres from the protagonist. The irregular faceted rock in the foreground covers independently moving adult participant 2 from about 2.8 s to 3.8 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 2.8 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.4 s to 1.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..382f420c6efd48b63852412b6da5c5219a170179
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-098-fantasy-gate/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the sliding door), <Subject 4> (the switch); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A massive, flat sliding door made of smooth, unadorned sandstone, set flush within the dark, recessed opening of the ruin. A small, rectangular control panel with a metallic finish, mounted on the wall to the right of the sliding door. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the sliding door; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the switch; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> pressing a control to open a sliding gate and passing through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A rocky mountain valley with rising paths, ledges, steep rock walls and a narrow gorge. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A massive, flat sliding door made of smooth, unadorned sandstone, set flush within the dark, recessed opening of the ruin. A small, rectangular control panel with a metallic finish, mounted on the wall to the right of the sliding door. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing centrally in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the sliding door, realized with the materials above. <Subject 4> is the switch, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a park ranger with short grey hair under a wide-brim hat, wearing an olive ranger shirt, tan trousers, and brown boots, standing to the left of the central figure - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing to the far right of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back on the far left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the sliding gate and its control; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:00.967 to 00:02.067, the protagonist presses the control and the gate slides open; the protagonist moves about a metre along its route, lowers the hands to the sides; the door moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open gate and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist opens the sliding door from 00:00.000 to 00:02.067; the protagonist passed through the sliding door from 00:02.067 to 00:05.000. The sliding door is one continuous object throughout and comes to rest about 5 metres from the protagonist. The irregular faceted rock in the foreground covers independently moving adult participant 2 from about 2.8 s to 3.8 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 2.8 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.4 s to 1.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..011631b6b1dd081772d5bbf1dd77397a3fa8441e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust-red body, complete with doors, windows, and wheels, positioned in the middle right area, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. A vintage sedan with a slate-grey body, complete with doors, windows, and wheels, positioned in the upper centre background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is crossing a road as approaching vehicles brake and yield, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust-red body, complete with doors, windows, and wheels, positioned in the middle right area, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. A vintage sedan with a slate-grey body, complete with doors, windows, and wheels, positioned in the upper centre background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same palette-knife oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, positioned to the middle left, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the middle right, seen only from behind, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, located in the upper left background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks across the road while the approaching vehicles brake and yield; the protagonist moves about 10 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 7 metres along its route; the vehicle moves about 7 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the far side of the road and keeps walking; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the vehicle moves about half a metre along its route. <Subject 3> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The passenger car in the foreground covers independently moving adult participant 3 from about 1.5 s to 2.2 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d5b9bb24b9fb6600e3e3a747b6832ea12d27c4b3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust-red body, complete with doors, windows, and wheels, positioned in the middle right area, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. A vintage sedan with a slate-grey body, complete with doors, windows, and wheels, positioned in the upper centre background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is crossing a road as approaching vehicles brake and yield, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust-red body, complete with doors, windows, and wheels, positioned in the middle right area, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. A vintage sedan with a slate-grey body, complete with doors, windows, and wheels, positioned in the upper centre background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same palette-knife oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, positioned to the middle left, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the middle right, seen only from behind, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, located in the upper left background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks across the road while the approaching vehicles brake and yield; the protagonist moves about 10 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 7 metres along its route; the vehicle moves about 7 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the far side of the road and keeps walking; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the vehicle moves about half a metre along its route. <Subject 3> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The passenger car in the foreground covers independently moving adult participant 3 from about 1.5 s to 2.2 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..28ed02fd2e2c401f6a90aaa938a693ebe9a21c86
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-101-city-crossing/first-person/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust-red body, complete with doors, windows, and wheels, positioned in the middle right area, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. A vintage sedan with a slate-grey body, complete with doors, windows, and wheels, positioned in the upper centre background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is crossing a road as approaching vehicles brake and yield, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust-red body, complete with doors, windows, and wheels, positioned in the middle right area, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front away from the camera. A vintage sedan with a slate-grey body, complete with doors, windows, and wheels, positioned in the upper centre background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same palette-knife oil painting manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, positioned to the middle left, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the middle right, seen only from behind, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, located in the upper left background, painted in the same palette-knife oil painting manner (character design: figures as bright colour patches against a darker ground) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks across the road while the approaching vehicles brake and yield; the protagonist moves about 10 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 7 metres along its route; the vehicle moves about 7 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the far side of the road and keeps walking; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the vehicle moves about half a metre along its route. <Subject 3> is one continuous object throughout and ends the clip about 6 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The passenger car in the foreground covers independently moving adult participant 3 from about 1.5 s to 2.2 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, palette-knife oil painting: broad flat knife strokes, thick ridged paint, bold colour blocks, scraped textures and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a warm amber-and-teal palette, harsh midday sun with crisp shadows, still humid air after rain, a dark neutral ground behind the subjects; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-102-city-drive/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-102-city-drive/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c277d1939373c09545d41ab0890954c2beaaa779
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-102-city-drive/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek small hover-car with a matte slate body, hovering slightly above the ground on the right side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. A small hover-car with a matte slate body, hovering slightly above the ground on the left side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #a7f736 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the object proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> approaching a car, getting in and driving forward, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek small hover-car with a matte slate body, hovering slightly above the ground on the right side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. A small hover-car with a matte slate body, hovering slightly above the ground on the left side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the center foreground, seen only from behind, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing further back in the center, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing further back on the right, seen only from behind, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back in the center, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.067, the protagonist walks up to the parked car; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.2 m), raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:01.067 to 00:02.167, the protagonist opens the car door and ducks inside; the protagonist moves about one and a half metres along its route, bends the knees; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:02.167 to 00:02.633, the protagonist gets into the car and takes the driving seat; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route. From 00:02.633 to 00:05.000, the protagonist drives the car forward; the protagonist moves about 5 metres along its route; the vehicle moves about 5 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist occupies <Subject 3> from 00:02.167 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 5 metres from the protagonist. The passenger car in the foreground covers the protagonist from about 0.4 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 1 from about 2.8 s to 3.4 s; the passenger car in the foreground covers independently moving adult participant 3 from about 1.3 s to 2.1 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-102-city-drive/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-102-city-drive/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..65165261da423a09af82e1bcd0adfcd7dffa14f5
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-102-city-drive/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek small hover-car with a matte slate body, hovering slightly above the ground on the right side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. A small hover-car with a matte slate body, hovering slightly above the ground on the left side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> approaching a car, getting in and driving forward, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek small hover-car with a matte slate body, hovering slightly above the ground on the right side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. A small hover-car with a matte slate body, hovering slightly above the ground on the left side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the center foreground, seen only from behind, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing further back in the center, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing further back on the right, seen only from behind, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back in the center, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.067, the protagonist walks up to the parked car; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.2 m), raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:01.067 to 00:02.167, the protagonist opens the car door and ducks inside; the protagonist moves about one and a half metres along its route, bends the knees; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:02.167 to 00:02.633, the protagonist gets into the car and takes the driving seat; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route. From 00:02.633 to 00:05.000, the protagonist drives the car forward; the protagonist moves about 5 metres along its route; the vehicle moves about 5 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist occupies <Subject 3> from 00:02.167 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 5 metres from the protagonist. The passenger car in the foreground covers the protagonist from about 0.4 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 1 from about 2.8 s to 3.4 s; the passenger car in the foreground covers independently moving adult participant 3 from about 1.3 s to 2.1 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-102-city-drive/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-102-city-drive/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c3ec28e89acfc9e8ef783bdd4e035a38a37bedc3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-102-city-drive/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek small hover-car with a matte slate body, hovering slightly above the ground on the right side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. A small hover-car with a matte slate body, hovering slightly above the ground on the left side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> approaching a car, getting in and driving forward, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek small hover-car with a matte slate body, hovering slightly above the ground on the right side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. A small hover-car with a matte slate body, hovering slightly above the ground on the left side, its interior cavity visible, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the center foreground, seen only from behind, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing further back in the center, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing further back on the right, seen only from behind, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back in the center, painted in the same classical glazed oil painting manner (character design: figures blended softly into the light) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.067, the protagonist walks up to the parked car; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.2 m), raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:01.067 to 00:02.167, the protagonist opens the car door and ducks inside; the protagonist moves about one and a half metres along its route, bends the knees; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:02.167 to 00:02.633, the protagonist gets into the car and takes the driving seat; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route. From 00:02.633 to 00:05.000, the protagonist drives the car forward; the protagonist moves about 5 metres along its route; the vehicle moves about 5 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route. Recorded relations: the protagonist occupies <Subject 3> from 00:02.167 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 5 metres from the protagonist. The passenger car in the foreground covers the protagonist from about 0.4 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 1 from about 2.8 s to 3.4 s; the passenger car in the foreground covers independently moving adult participant 3 from about 1.3 s to 2.1 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, classical oil painting with smooth layered glazes: soft blended edges, deep warm chiaroscuro, varnished sheen and canvas weave instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, misty morning backlight, fine rain and wet reflective ground, long fluid strokes following each form, a background of loose suggestive strokes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-104-city-stairs/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-104-city-stairs/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d27d2c6bffdfd645ce51de158ac9c91323a32ffb
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-104-city-stairs/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust body shade, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #36f796 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36aff7 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the human proxy tagged #e7f736 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the human proxy tagged #ce36f7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust body shade, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 7 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The passenger car is one continuous object throughout and ends the clip about 12 metres from the protagonist. The platform-rail-1 in the foreground covers independently moving adult participant 3 from about 3.1 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-104-city-stairs/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-104-city-stairs/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4e9527d21f0c008e0413980e1b23b2498e6ed258
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-104-city-stairs/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust body shade, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust body shade, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 7 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The passenger car is one continuous object throughout and ends the clip about 12 metres from the protagonist. The platform-rail-1 in the foreground covers independently moving adult participant 3 from about 3.1 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-104-city-stairs/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-104-city-stairs/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d085334fa59f97862fadfab8f8049aaadc1aeffd
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-104-city-stairs/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (independently moving adult participant 3) and 1 object proxy standing for <Subject 3> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust body shade, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> climbing stairs and reaching the upper landing, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A vintage sedan with a rust body shade, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> (independently moving adult participant 1) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 3) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist climbs the stairs step by step; the protagonist moves about 5 metres along its route, rises about a metre; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the vehicle moves about 7 metres along its route. From 00:04.667 to 00:05.000, the protagonist reaches the upper landing and walks on; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The passenger car is one continuous object throughout and ends the clip about 12 metres from the protagonist. The platform-rail-1 in the foreground covers independently moving adult participant 3 from about 3.1 s to 3.8 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-105-city-door/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-105-city-door/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..47138d13eff14cf0e8db51c9e26df1fd40115444
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-105-city-door/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the hinged door), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 1.3 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular hinged door set into the building facade, rendered with precise ruled edges and light construction lines to indicate its flat, wooden surface, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors). A vintage sedan positioned on the right side of the frame, its body shaded in cool grey graphite with detailed hatching to suggest a metallic finish, complete with visible doors, windows, and wheels, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the hinged door; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the object proxy tagged #6636f7 that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2: not in frame at 00:00.000; it enters the frame at about 1.3 s; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> opening a hinged door and walking through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (enters [Shot 1] at about 1.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular hinged door set into the building facade, rendered with precise ruled edges and light construction lines to indicate its flat, wooden surface, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors). A vintage sedan positioned on the right side of the frame, its body shaded in cool grey graphite with detailed hatching to suggest a metallic finish, complete with visible doors, windows, and wheels, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the frame, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the hinged door, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing on the far left edge of the frame, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is not in frame at 00:00.000 and enters the frame at about 1.3 s; from then on it is a mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, positioned in the upper center background, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing in the upper left background, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the hinged door; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route. From 00:00.967 to 00:02.067, the protagonist swings the door open; the protagonist moves about a metre along its route, lowers the hands to the sides; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about 2 metres along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open doorway and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 6 metres along its route. Recorded relations: the protagonist opens the hinged door from 00:00.000 to 00:02.067; the protagonist passed through the hinged door from 00:02.067 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 10 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 3.8 s to 4.4 s; the individual street building in the foreground covers independently moving adult participant 1 from about 2.5 s to 4.6 s; the door-frame--1 in the foreground covers independently moving adult participant 1 from about 0.2 s to 0.9 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.5 s to 2.2 s; the hinged door in the foreground covers independently moving adult participant 2 from about 0.6 s to 1.4 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.6 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 2.7 s to 4.3 s; the hinged door in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.8 s; the hinged door in the foreground covers independently moving adult participant 3 from about 2.4 s to 3.1 s; the door-frame--1 in the foreground covers independently moving adult participant 3 from about 0.9 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-105-city-door/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-105-city-door/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4eb1fb2f36668d0f041ea8981c5c61e7c76750f3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-105-city-door/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the hinged door), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 1.3 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular hinged door set into the building facade, rendered with precise ruled edges and light construction lines to indicate its flat, wooden surface, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors). A vintage sedan positioned on the right side of the frame, its body shaded in cool grey graphite with detailed hatching to suggest a metallic finish, complete with visible doors, windows, and wheels, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the hinged door; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2: not in frame at 00:00.000; it enters the frame at about 1.3 s; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> opening a hinged door and walking through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (enters [Shot 1] at about 1.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular hinged door set into the building facade, rendered with precise ruled edges and light construction lines to indicate its flat, wooden surface, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors). A vintage sedan positioned on the right side of the frame, its body shaded in cool grey graphite with detailed hatching to suggest a metallic finish, complete with visible doors, windows, and wheels, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the frame, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the hinged door, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing on the far left edge of the frame, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is not in frame at 00:00.000 and enters the frame at about 1.3 s; from then on it is a mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, positioned in the upper center background, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing in the upper left background, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the hinged door; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route. From 00:00.967 to 00:02.067, the protagonist swings the door open; the protagonist moves about a metre along its route, lowers the hands to the sides; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about 2 metres along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open doorway and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 6 metres along its route. Recorded relations: the protagonist opens the hinged door from 00:00.000 to 00:02.067; the protagonist passed through the hinged door from 00:02.067 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 10 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 3.8 s to 4.4 s; the individual street building in the foreground covers independently moving adult participant 1 from about 2.5 s to 4.6 s; the door-frame--1 in the foreground covers independently moving adult participant 1 from about 0.2 s to 0.9 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.5 s to 2.2 s; the hinged door in the foreground covers independently moving adult participant 2 from about 0.6 s to 1.4 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.6 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 2.7 s to 4.3 s; the hinged door in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.8 s; the hinged door in the foreground covers independently moving adult participant 3 from about 2.4 s to 3.1 s; the door-frame--1 in the foreground covers independently moving adult participant 3 from about 0.9 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-105-city-door/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-105-city-door/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1a605393291216b5584982011cce901bef7c4c23
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-105-city-door/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the hinged door), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 1.3 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular hinged door set into the building facade, rendered with precise ruled edges and light construction lines to indicate its flat, wooden surface, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors). A vintage sedan positioned on the right side of the frame, its body shaded in cool grey graphite with detailed hatching to suggest a metallic finish, complete with visible doors, windows, and wheels, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the hinged door; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2: not in frame at 00:00.000; it enters the frame at about 1.3 s; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> opening a hinged door and walking through, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (enters [Shot 1] at about 1.3 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large, rectangular hinged door set into the building facade, rendered with precise ruled edges and light construction lines to indicate its flat, wooden surface, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors). A vintage sedan positioned on the right side of the frame, its body shaded in cool grey graphite with detailed hatching to suggest a metallic finish, complete with visible doors, windows, and wheels, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the frame, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the hinged door, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing on the far left edge of the frame, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is not in frame at 00:00.000 and enters the frame at about 1.3 s; from then on it is a mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, positioned in the upper center background, seen only from behind, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing in the upper left background, drawn in the same design-marker render sketch manner (character design: figures with dark heavy outlines and light interiors) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the hinged door; the protagonist moves about one and a half metres along its route, raises the hands to chest height; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route. From 00:00.967 to 00:02.067, the protagonist swings the door open; the protagonist moves about a metre along its route, lowers the hands to the sides; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about 2 metres along its route. From 00:02.067 to 00:05.000, the protagonist walks through the open doorway and continues beyond it; the protagonist moves about 3 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the vehicle moves about 6 metres along its route. Recorded relations: the protagonist opens the hinged door from 00:00.000 to 00:02.067; the protagonist passed through the hinged door from 00:02.067 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 10 metres from the protagonist. The individual street building in the foreground covers independently moving adult participant 1 from about 3.8 s to 4.4 s; the individual street building in the foreground covers independently moving adult participant 1 from about 2.5 s to 4.6 s; the door-frame--1 in the foreground covers independently moving adult participant 1 from about 0.2 s to 0.9 s; the protagonist in the foreground covers independently moving adult participant 2 from about 0.5 s to 2.2 s; the hinged door in the foreground covers independently moving adult participant 2 from about 0.6 s to 1.4 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.6 s; the door-frame-1 in the foreground covers independently moving adult participant 2 from about 2.7 s to 4.3 s; the hinged door in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.8 s; the hinged door in the foreground covers independently moving adult participant 3 from about 2.4 s to 3.1 s; the door-frame--1 in the foreground covers independently moving adult participant 3 from about 0.9 s to 2.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, design-marker sketch look: fine ink linework with broad sheer marker fills in cool greys and one accent colour, streaky marker overlaps, bright white paper, drawn over the full 3D world so every recorded shape stays readable; a rich saturated jewel-tone palette, blue-hour dusk with a few warm practical lights, still humid air after rain, a few construction lines left visible; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4959be3b9df36cb4cb6e75d3d95c32de0e3ebdfa
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/colored-obb/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 human proxies standing for <Subject 4> (the person receiving the object), <Subject 5> (the independently moving adult participant) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 3> is not in frame at 00:00.000 and enters at about 1.5 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A rectangular wooden crate with a rough, weathered finish and visible grain, held by the protagonist. A dark grey pickup truck with a glossy finish, featuring a visible open cargo bed, side mirrors, and four wheels - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate: not in frame at 00:00.000; it enters the frame at about 1.5 s; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the independently moving adult participant; in <Video 1> the human proxy tagged #f79e36 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (enters [Shot 1] at about 1.5 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A rectangular wooden crate with a rough, weathered finish and visible grain, held by the protagonist. A dark grey pickup truck with a glossy finish, featuring a visible open cargo bed, side mirrors, and four wheels - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, and keeps exactly this appearance. <Subject 5> (the independently moving adult participant) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; the independently moving adult participant moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; the independently moving adult participant moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; the independently moving adult participant moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; the independently moving adult participant moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9f2d2fbb8e09af8ad6545ce45282b14c1da6c112
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/mixed/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 blue mannequin proxies standing for <Subject 4> (the person receiving the object), <Subject 5> (the independently moving adult participant) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 3> is not in frame at 00:00.000 and enters at about 1.5 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A rectangular wooden crate with a rough, weathered finish and visible grain, held by the protagonist. A dark grey pickup truck with a glossy finish, featuring a visible open cargo bed, side mirrors, and four wheels - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate: not in frame at 00:00.000; it enters the frame at about 1.5 s; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 5> is the independently moving adult participant; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (enters [Shot 1] at about 1.5 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A rectangular wooden crate with a rough, weathered finish and visible grain, held by the protagonist. A dark grey pickup truck with a glossy finish, featuring a visible open cargo bed, side mirrors, and four wheels - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, and keeps exactly this appearance. <Subject 5> (the independently moving adult participant) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; the independently moving adult participant moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; the independently moving adult participant moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; the independently moving adult participant moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; the independently moving adult participant moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bfe85a79cff964a8299e2738b667ff09ae6b2d38
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-106-city-handoff/first-person/white/prompt.txt
@@ -0,0 +1,40 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 grey block-figure proxies standing for <Subject 4> (the person receiving the object), <Subject 5> (the independently moving adult participant) and 1 object proxy standing for <Subject 3> (the parcel or cargo crate); nothing else in <Video 1> is a person or an interaction object. <Subject 3> is not in frame at 00:00.000 and enters at about 1.5 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A rectangular wooden crate with a rough, weathered finish and visible grain, held by the protagonist. A dark grey pickup truck with a glossy finish, featuring a visible open cargo bed, side mirrors, and four wheels - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate: not in frame at 00:00.000; it enters the frame at about 1.5 s; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 5> is the independently moving adult participant; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is offering a parcel to another person and releasing it to the receiver, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (enters [Shot 1] at about 1.5 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A rectangular wooden crate with a rough, weathered finish and visible grain, held by the protagonist. A dark grey pickup truck with a glossy finish, featuring a visible open cargo bed, side mirrors, and four wheels - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> (the person receiving the object) is an office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, and keeps exactly this appearance. <Subject 5> (the independently moving adult participant) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; the independently moving adult participant moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; the independently moving adult participant moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; the independently moving adult participant moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; the independently moving adult participant moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-108-city-cart/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-108-city-cart/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4a6bdf70ab591297f1ba18169b60c01b0f5b69fa
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-108-city-cart/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the wheeled handcart), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A second wheeled handcart, similar in design to the first, located in the middle right of the frame, slightly further away - at the start of the clip its front toward the camera. A wheeled handcart with a wooden frame and metal wheels, positioned in the lower centre of the frame, appearing sturdy and utilitarian - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> pushing a handcart forward around a gentle bend and releasing the handle, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A second wheeled handcart, similar in design to the first, located in the middle right of the frame, slightly further away - at the start of the clip its front toward the camera. A wheeled handcart with a wooden frame and metal wheels, positioned in the lower centre of the frame, appearing sturdy and utilitarian - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a street vendor with grey-streaked black hair, wearing a khaki apron over a sage tee, tan chinos, and rubber clogs, standing in the middle right of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle far right of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle right of the frame, slightly behind the street vendor - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist steps up to the handcart and takes hold of its handle; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:00.967 to 00:04.067, the protagonist pushes the handcart forward around the gentle bend; the protagonist moves about 5 metres along its route; the cart moves about 5 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.067 to 00:05.000, the protagonist lets go of the handle and the handcart stays where it stopped; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist pushes <Subject 3> from 00:00.967 to 00:04.067. <Subject 3> is one continuous object throughout and comes to rest about a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 3.1 s to 5.0 s; the wheeled handcart in the foreground covers independently moving adult participant 1 from about 3.1 s to 3.4 s; the protagonist in the foreground covers independently moving adult participant 3 from about 2.7 s to 3.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-108-city-cart/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-108-city-cart/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..883707608e59e0f4e4e485e9c455419f62986d4b
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-108-city-cart/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the wheeled handcart), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A second wheeled handcart, similar in design to the first, located in the middle right of the frame, slightly further away - at the start of the clip its front toward the camera. A wheeled handcart with a wooden frame and metal wheels, positioned in the lower centre of the frame, appearing sturdy and utilitarian - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> pushing a handcart forward around a gentle bend and releasing the handle, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A second wheeled handcart, similar in design to the first, located in the middle right of the frame, slightly further away - at the start of the clip its front toward the camera. A wheeled handcart with a wooden frame and metal wheels, positioned in the lower centre of the frame, appearing sturdy and utilitarian - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a street vendor with grey-streaked black hair, wearing a khaki apron over a sage tee, tan chinos, and rubber clogs, standing in the middle right of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle far right of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle right of the frame, slightly behind the street vendor - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist steps up to the handcart and takes hold of its handle; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:00.967 to 00:04.067, the protagonist pushes the handcart forward around the gentle bend; the protagonist moves about 5 metres along its route; the cart moves about 5 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.067 to 00:05.000, the protagonist lets go of the handle and the handcart stays where it stopped; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist pushes <Subject 3> from 00:00.967 to 00:04.067. <Subject 3> is one continuous object throughout and comes to rest about a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 3.1 s to 5.0 s; the wheeled handcart in the foreground covers independently moving adult participant 1 from about 3.1 s to 3.4 s; the protagonist in the foreground covers independently moving adult participant 3 from about 2.7 s to 3.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-108-city-cart/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-108-city-cart/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..736738a94efe7e4b739464fd7075f0faeb0d406a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-108-city-cart/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the wheeled handcart), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A second wheeled handcart, similar in design to the first, located in the middle right of the frame, slightly further away - at the start of the clip its front toward the camera. A wheeled handcart with a wooden frame and metal wheels, positioned in the lower centre of the frame, appearing sturdy and utilitarian - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> pushing a handcart forward around a gentle bend and releasing the handle, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A second wheeled handcart, similar in design to the first, located in the middle right of the frame, slightly further away - at the start of the clip its front toward the camera. A wheeled handcart with a wooden frame and metal wheels, positioned in the lower centre of the frame, appearing sturdy and utilitarian - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a street vendor with grey-streaked black hair, wearing a khaki apron over a sage tee, tan chinos, and rubber clogs, standing in the middle right of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle far right of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle right of the frame, slightly behind the street vendor - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist steps up to the handcart and takes hold of its handle; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:00.967 to 00:04.067, the protagonist pushes the handcart forward around the gentle bend; the protagonist moves about 5 metres along its route; the cart moves about 5 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.067 to 00:05.000, the protagonist lets go of the handle and the handcart stays where it stopped; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist pushes <Subject 3> from 00:00.967 to 00:04.067. <Subject 3> is one continuous object throughout and comes to rest about a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 3.1 s to 5.0 s; the wheeled handcart in the foreground covers independently moving adult participant 1 from about 3.1 s to 3.4 s; the protagonist in the foreground covers independently moving adult participant 3 from about 2.7 s to 3.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-109-city-carry/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-109-city-carry/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d7a58a1cf91ae555f87c3fcf6ce381985aadea4c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-109-city-carry/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal and wood, positioned further back in the market aisle - at the start of the clip its front toward the camera. A sturdy wooden crate with a matte finish and visible grain, supported by the pickup stand behind the protagonist. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal and wood, positioned further back in the market aisle - at the start of the clip its front toward the camera. A sturdy wooden crate with a matte finish and visible grain, supported by the pickup stand behind the protagonist. <Subject 1>, the protagonist, has the following target appearance: A realistic adult street vendor with grey-streaked black hair, wearing a khaki apron, tan chinos, and rubber clogs, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 0.7 s to 1.8 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.8 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 2.4 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-109-city-carry/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-109-city-carry/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fa494ad2c0da4c6aa1dbaa671a905fd993beaffb
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-109-city-carry/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal and wood, positioned further back in the market aisle - at the start of the clip its front toward the camera. A sturdy wooden crate with a matte finish and visible grain, supported by the pickup stand behind the protagonist. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal and wood, positioned further back in the market aisle - at the start of the clip its front toward the camera. A sturdy wooden crate with a matte finish and visible grain, supported by the pickup stand behind the protagonist. <Subject 1>, the protagonist, has the following target appearance: A realistic adult street vendor with grey-streaked black hair, wearing a khaki apron, tan chinos, and rubber clogs, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 0.7 s to 1.8 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.8 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 2.4 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-109-city-carry/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-109-city-carry/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..823e18e59045ab96810b8ed0f456e9df94722875
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-109-city-carry/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal and wood, positioned further back in the market aisle - at the start of the clip its front toward the camera. A sturdy wooden crate with a matte finish and visible grain, supported by the pickup stand behind the protagonist. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A narrow old-town market with stone facades, stalls, awnings and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of dark metal and wood, positioned further back in the market aisle - at the start of the clip its front toward the camera. A sturdy wooden crate with a matte finish and visible grain, supported by the pickup stand behind the protagonist. <Subject 1>, the protagonist, has the following target appearance: A realistic adult street vendor with grey-streaked black hair, wearing a khaki apron, tan chinos, and rubber clogs, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the distance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 3 from about 0.7 s to 1.8 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.8 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 2.4 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-110-city-throw/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-110-city-throw/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4e707cfecb3c1097eb41266640e3905a613e219a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-110-city-throw/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person catching the ball), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the throwable ball), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek, chrome handcart with a polished finish, reflecting the surrounding neon lights, positioned near the center - at the start of the clip its front toward the camera. A glowing, clear sphere with a soft, pulsating light, hovering slightly above the ground. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the throwable ball; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #3ef736; use the target appearance described below.
+<Subject 5> is the person catching the ball; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #6636f7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> throwing a ball through the air to a second person who catches it, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek, chrome handcart with a polished finish, reflecting the surrounding neon lights, positioned near the center - at the start of the clip its front toward the camera. A glowing, clear sphere with a soft, pulsating light, hovering slightly above the ground. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the throwable ball, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person catching the ball) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, ready to catch the ball - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist winds up and throws the ball toward the second person; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.267 to 00:02.367, the ball flies through the air toward the second person; the protagonist lowers the hands; the ball moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:02.367 to 00:05.000, the second person catches the ball and holds it; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. Recorded relations: the person catching the ball holds the throwable ball from 00:02.367 to 00:05.000. The throwable ball is one continuous object throughout and comes to rest about 4 metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The protagonist in the foreground covers the person catching the ball from about 1.5 s to 3.7 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 3.2 s to 4.3 s; the protagonist in the foreground covers independently moving adult participant 3 from about 0.9 s to 1.2 s; the protagonist in the foreground covers independently moving adult participant 3 from about 4.2 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-110-city-throw/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-110-city-throw/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c584d6f13c51a9b6d9c808c5cd0cc39d48afc6a1
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-110-city-throw/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person catching the ball), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the throwable ball), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek, chrome handcart with a polished finish, reflecting the surrounding neon lights, positioned near the center - at the start of the clip its front toward the camera. A glowing, clear sphere with a soft, pulsating light, hovering slightly above the ground. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the throwable ball; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 5> is the person catching the ball; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> throwing a ball through the air to a second person who catches it, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek, chrome handcart with a polished finish, reflecting the surrounding neon lights, positioned near the center - at the start of the clip its front toward the camera. A glowing, clear sphere with a soft, pulsating light, hovering slightly above the ground. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the throwable ball, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person catching the ball) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, ready to catch the ball - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist winds up and throws the ball toward the second person; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.267 to 00:02.367, the ball flies through the air toward the second person; the protagonist lowers the hands; the ball moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:02.367 to 00:05.000, the second person catches the ball and holds it; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. Recorded relations: the person catching the ball holds the throwable ball from 00:02.367 to 00:05.000. The throwable ball is one continuous object throughout and comes to rest about 4 metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The protagonist in the foreground covers the person catching the ball from about 1.5 s to 3.7 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 3.2 s to 4.3 s; the protagonist in the foreground covers independently moving adult participant 3 from about 0.9 s to 1.2 s; the protagonist in the foreground covers independently moving adult participant 3 from about 4.2 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-110-city-throw/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-110-city-throw/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..84d4a9302790df8304feb8a719f288688927d0d5
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-110-city-throw/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person catching the ball), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the throwable ball), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek, chrome handcart with a polished finish, reflecting the surrounding neon lights, positioned near the center - at the start of the clip its front toward the camera. A glowing, clear sphere with a soft, pulsating light, hovering slightly above the ground. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the throwable ball; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the person catching the ball; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> throwing a ball through the air to a second person who catches it, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A sleek, chrome handcart with a polished finish, reflecting the surrounding neon lights, positioned near the center - at the start of the clip its front toward the camera. A glowing, clear sphere with a soft, pulsating light, hovering slightly above the ground. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the throwable ball, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person catching the ball) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, ready to catch the ball - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is an office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing further back - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.267, the protagonist winds up and throws the ball toward the second person; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.267 to 00:02.367, the ball flies through the air toward the second person; the protagonist lowers the hands; the ball moves about 3 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:02.367 to 00:05.000, the second person catches the ball and holds it; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. Recorded relations: the person catching the ball holds the throwable ball from 00:02.367 to 00:05.000. The throwable ball is one continuous object throughout and comes to rest about 4 metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The protagonist in the foreground covers the person catching the ball from about 1.5 s to 3.7 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 1 from about 3.2 s to 4.3 s; the protagonist in the foreground covers independently moving adult participant 3 from about 0.9 s to 1.2 s; the protagonist in the foreground covers independently moving adult participant 3 from about 4.2 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-111-city-handoff/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-111-city-handoff/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..751f808962c5f9e96fbb99ad1dc8c3402e059130
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-111-city-handoff/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A black metal handcart with a simple frame and rubber wheels, standing empty on the paved path - at the start of the clip its front toward the camera. A sturdy wooden crate with visible grain and iron corner reinforcements, held by the protagonist with a weathered finish. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #3ef736; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #6636f7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A black metal handcart with a simple frame and rubber wheels, standing empty on the paved path - at the start of the clip its front toward the camera. A sturdy wooden crate with visible grain and iron corner reinforcements, held by the protagonist with a weathered finish. <Subject 1>, the protagonist, has the following target appearance: A youthful student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing attentively, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, walking along the path - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, strolling leisurely, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing near a bench - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about one and a half metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 3.4 s to 4.4 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 2 from about 3.4 s to 4.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-111-city-handoff/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-111-city-handoff/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4b73274116fb6ff527064bdba3466d4abcb8779f
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-111-city-handoff/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A black metal handcart with a simple frame and rubber wheels, standing empty on the paved path - at the start of the clip its front toward the camera. A sturdy wooden crate with visible grain and iron corner reinforcements, held by the protagonist with a weathered finish. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A black metal handcart with a simple frame and rubber wheels, standing empty on the paved path - at the start of the clip its front toward the camera. A sturdy wooden crate with visible grain and iron corner reinforcements, held by the protagonist with a weathered finish. <Subject 1>, the protagonist, has the following target appearance: A youthful student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing attentively, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, walking along the path - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, strolling leisurely, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing near a bench - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about one and a half metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 3.4 s to 4.4 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 2 from about 3.4 s to 4.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-111-city-handoff/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-111-city-handoff/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f32aa7068df4433173c547a3bb990baeabe00b5d
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-111-city-handoff/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A black metal handcart with a simple frame and rubber wheels, standing empty on the paved path - at the start of the clip its front toward the camera. A sturdy wooden crate with visible grain and iron corner reinforcements, held by the protagonist with a weathered finish. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A black metal handcart with a simple frame and rubber wheels, standing empty on the paved path - at the start of the clip its front toward the camera. A sturdy wooden crate with visible grain and iron corner reinforcements, held by the protagonist with a weathered finish. <Subject 1>, the protagonist, has the following target appearance: A youthful student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing with a relaxed posture; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing attentively, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, walking along the path - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, strolling leisurely, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing near a bench - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about one and a half metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 3.4 s to 4.4 s; the tree with a distinct trunk and crown in the foreground covers independently moving adult participant 2 from about 3.4 s to 4.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-113-city-tow/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-113-city-tow/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..02a2081d46d4ee6fe0674fc8f7448ed5e534cfc8
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-113-city-tow/follow/colored-obb/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 7> (independently moving adult participant 1), <Subject 8> (independently moving adult participant 2), <Subject 9> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car), <Subject 5> (the tow connection), <Subject 6> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A weathered grey pickup truck with an open bed, positioned in the lower right, its body showing signs of wear and rust, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A faded maroon vintage sedan with a rounded body, positioned in the lower far right, its paint chipped and faded, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A white delivery van with a boxy shape, positioned in the lower left, its surface clean but showing minor scuffs, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #a7f736 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the object proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the tow connection: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the cargo utility truck; in <Video 1> the object proxy tagged #f73677 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+<Subject 8> is independently moving adult participant 2; in <Video 1> the human proxy tagged #6636f7 that starts on the right of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 3; in <Video 1> the human proxy tagged #3ef736; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> entering a tow vehicle, connecting a second car and towing it forward, involving <Subject 3> to <Subject 9>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A weathered grey pickup truck with an open bed, positioned in the lower right, its body showing signs of wear and rust, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A faded maroon vintage sedan with a rounded body, positioned in the lower far right, its paint chipped and faded, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A white delivery van with a boxy shape, positioned in the lower left, its surface clean but showing minor scuffs, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the middle ground, seen only from behind, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> is the tow connection, realized with the materials above. <Subject 6> is the cargo utility truck, realized with the materials above. <Subject 7> (independently moving adult participant 1) is an adult male construction worker with short black hair under a hard hat, wearing an amber high-visibility vest over a denim shirt, khaki work trousers, and tan work boots, standing in the upper centre, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 2) is an adult male security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in the upper right, seen only from behind, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 3) is an adult male bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the upper centre, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.067, the protagonist walks up to the parked tow vehicle; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.2 m), raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:01.067 to 00:02.167, the protagonist opens the tow vehicle's door and ducks inside; the protagonist moves about one and a half metres along its route, bends the knees; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about a metre along its route. From 00:02.167 to 00:02.367, the protagonist gets into the tow vehicle. From 00:02.367 to 00:02.733, the tow connection links the tow vehicle to the second car; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route. From 00:02.733 to 00:05.000, the tow vehicle pulls the second car forward with the protagonist at the wheel; the protagonist moves about 4 metres along its route; the vehicle moves about 4 metres along its route; the vehicle moves about 4 metres along its route; the cable moves about 4 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the protagonist occupies <Subject 3> from 00:02.167 to 00:05.000; <Subject 3> tows <Subject 4> from 00:02.367 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 5 metres from the protagonist. The tow connection is one continuous object throughout and ends the clip about half a metre from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 4 metres from the protagonist. The passenger car in the foreground covers the protagonist from about 0.5 s to 5.0 s; the passenger car in the foreground covers the protagonist from about 1.2 s to 1.4 s; the passenger car in the foreground covers the protagonist from about 1.4 s to 1.6 s; the passenger car in the foreground covers the protagonist from about 4.5 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 1 from about 1.1 s to 4.6 s; the passenger car in the foreground covers independently moving adult participant 1 from about 4.2 s to 4.6 s; the cargo utility truck in the foreground covers independently moving adult participant 1 from about 4.8 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 3 from about 3.5 s to 3.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-113-city-tow/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-113-city-tow/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fa1d5c0ecc9c32a36a8eb3cf3f157aa761dd63d7
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-113-city-tow/follow/mixed/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 7> (independently moving adult participant 1), <Subject 8> (independently moving adult participant 2), <Subject 9> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car), <Subject 5> (the tow connection), <Subject 6> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A weathered grey pickup truck with an open bed, positioned in the lower right, its body showing signs of wear and rust, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A faded maroon vintage sedan with a rounded body, positioned in the lower far right, its paint chipped and faded, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A white delivery van with a boxy shape, positioned in the lower left, its surface clean but showing minor scuffs, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the tow connection: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the black solid cable proxy shape; use the target appearance described below.
+<Subject 6> is the cargo utility truck; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 8> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> entering a tow vehicle, connecting a second car and towing it forward, involving <Subject 3> to <Subject 9>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A weathered grey pickup truck with an open bed, positioned in the lower right, its body showing signs of wear and rust, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A faded maroon vintage sedan with a rounded body, positioned in the lower far right, its paint chipped and faded, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A white delivery van with a boxy shape, positioned in the lower left, its surface clean but showing minor scuffs, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the middle ground, seen only from behind, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> is the tow connection, realized with the materials above. <Subject 6> is the cargo utility truck, realized with the materials above. <Subject 7> (independently moving adult participant 1) is an adult male construction worker with short black hair under a hard hat, wearing an amber high-visibility vest over a denim shirt, khaki work trousers, and tan work boots, standing in the upper centre, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 2) is an adult male security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in the upper right, seen only from behind, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 3) is an adult male bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the upper centre, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.067, the protagonist walks up to the parked tow vehicle; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.2 m), raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:01.067 to 00:02.167, the protagonist opens the tow vehicle's door and ducks inside; the protagonist moves about one and a half metres along its route, bends the knees; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about a metre along its route. From 00:02.167 to 00:02.367, the protagonist gets into the tow vehicle. From 00:02.367 to 00:02.733, the tow connection links the tow vehicle to the second car; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route. From 00:02.733 to 00:05.000, the tow vehicle pulls the second car forward with the protagonist at the wheel; the protagonist moves about 4 metres along its route; the vehicle moves about 4 metres along its route; the vehicle moves about 4 metres along its route; the cable moves about 4 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the protagonist occupies <Subject 3> from 00:02.167 to 00:05.000; <Subject 3> tows <Subject 4> from 00:02.367 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 5 metres from the protagonist. The tow connection is one continuous object throughout and ends the clip about half a metre from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 4 metres from the protagonist. The passenger car in the foreground covers the protagonist from about 0.5 s to 5.0 s; the passenger car in the foreground covers the protagonist from about 1.2 s to 1.4 s; the passenger car in the foreground covers the protagonist from about 1.4 s to 1.6 s; the passenger car in the foreground covers the protagonist from about 4.5 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 1 from about 1.1 s to 4.6 s; the passenger car in the foreground covers independently moving adult participant 1 from about 4.2 s to 4.6 s; the cargo utility truck in the foreground covers independently moving adult participant 1 from about 4.8 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 3 from about 3.5 s to 3.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-113-city-tow/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-113-city-tow/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..887173163de16f8e02ec4966e908edc1d0c947e9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-113-city-tow/follow/white/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 7> (independently moving adult participant 1), <Subject 8> (independently moving adult participant 2), <Subject 9> (independently moving adult participant 3) and 4 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the passenger car), <Subject 5> (the tow connection), <Subject 6> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A weathered grey pickup truck with an open bed, positioned in the lower right, its body showing signs of wear and rust, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A faded maroon vintage sedan with a rounded body, positioned in the lower far right, its paint chipped and faded, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A white delivery van with a boxy shape, positioned in the lower left, its surface clean but showing minor scuffs, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is the tow connection: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the cargo utility truck; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 8> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 9> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> entering a tow vehicle, connecting a second car and towing it forward, involving <Subject 3> to <Subject 9>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An industrial work yard with steel gantries, scaffolding, cargo, platforms and vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A weathered grey pickup truck with an open bed, positioned in the lower right, its body showing signs of wear and rust, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A faded maroon vintage sedan with a rounded body, positioned in the lower far right, its paint chipped and faded, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. A white delivery van with a boxy shape, positioned in the lower left, its surface clean but showing minor scuffs, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult male site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the middle ground, seen only from behind, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> is the tow connection, realized with the materials above. <Subject 6> is the cargo utility truck, realized with the materials above. <Subject 7> (independently moving adult participant 1) is an adult male construction worker with short black hair under a hard hat, wearing an amber high-visibility vest over a denim shirt, khaki work trousers, and tan work boots, standing in the upper centre, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 2) is an adult male security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in the upper right, seen only from behind, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 9> (independently moving adult participant 3) is an adult male bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the upper centre, rendered as a stylized cartoon 3D character in the same toy-like cute 3D manner (character design: pear-shaped bodies with short legs) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.067, the protagonist walks up to the parked tow vehicle; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.2 m), raises the hands to chest height; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about one and a half metres along its route. From 00:01.067 to 00:02.167, the protagonist opens the tow vehicle's door and ducks inside; the protagonist moves about one and a half metres along its route, bends the knees; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about a metre along its route. From 00:02.167 to 00:02.367, the protagonist gets into the tow vehicle. From 00:02.367 to 00:02.733, the tow connection links the tow vehicle to the second car; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route. From 00:02.733 to 00:05.000, the tow vehicle pulls the second car forward with the protagonist at the wheel; the protagonist moves about 4 metres along its route; the vehicle moves about 4 metres along its route; the vehicle moves about 4 metres along its route; the cable moves about 4 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about 2 metres along its route. Recorded relations: the protagonist occupies <Subject 3> from 00:02.167 to 00:05.000; <Subject 3> tows <Subject 4> from 00:02.367 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 5 metres from the protagonist. The tow connection is one continuous object throughout and ends the clip about half a metre from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 4 metres from the protagonist. The passenger car in the foreground covers the protagonist from about 0.5 s to 5.0 s; the passenger car in the foreground covers the protagonist from about 1.2 s to 1.4 s; the passenger car in the foreground covers the protagonist from about 1.4 s to 1.6 s; the passenger car in the foreground covers the protagonist from about 4.5 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 1 from about 1.1 s to 4.6 s; the passenger car in the foreground covers independently moving adult participant 1 from about 4.2 s to 4.6 s; the cargo utility truck in the foreground covers independently moving adult participant 1 from about 4.8 s to 5.0 s; the passenger car in the foreground covers independently moving adult participant 3 from about 3.5 s to 3.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — toy-like cute 3D: rounded matte shapes, miniature proportions, pastel colours and soft even lighting; a sun-bleached faded palette, flat overcast daylight, light drifting haze, soft airbrushed shading, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-116-city-handoff/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-116-city-handoff/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..833deea57521f24649428e630472960823ed2a6d
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-116-city-handoff/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing). A rugged pickup truck with a slate-grey body, complete with doors, windows, and wheels, positioned on the left side, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the cargo utility truck; in <Video 1> the object proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the human proxy tagged #6636f7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing). A rugged pickup truck with a slate-grey body, complete with doors, windows, and wheels, positioned on the left side, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the foreground, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing); keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the cargo utility truck, realized with the materials above. <Subject 5> (the person receiving the object) is a stocky adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing opposite the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing), and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a stocky adult fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the background, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a stocky adult construction worker with short black hair, wearing an amber high-visibility vest, khaki work trousers, and tan work boots, standing in the background, seen only from behind, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a stocky adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the background, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 7 metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 4.4 s to 4.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-116-city-handoff/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-116-city-handoff/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f34feb59deba2a498e395175500da5cd17a1294f
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-116-city-handoff/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing). A rugged pickup truck with a slate-grey body, complete with doors, windows, and wheels, positioned on the left side, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the cargo utility truck; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing). A rugged pickup truck with a slate-grey body, complete with doors, windows, and wheels, positioned on the left side, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the foreground, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing); keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the cargo utility truck, realized with the materials above. <Subject 5> (the person receiving the object) is a stocky adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing opposite the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing), and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a stocky adult fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the background, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a stocky adult construction worker with short black hair, wearing an amber high-visibility vest, khaki work trousers, and tan work boots, standing in the background, seen only from behind, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a stocky adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the background, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 7 metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 4.4 s to 4.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-116-city-handoff/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-116-city-handoff/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..711762be8728f347ba0c65dbefac79e2c36e2b22
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-116-city-handoff/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the cargo utility truck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing). A rugged pickup truck with a slate-grey body, complete with doors, windows, and wheels, positioned on the left side, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the cargo utility truck; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 8> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wooden cargo crate with a weathered finish and metal corner reinforcements, held by the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing). A rugged pickup truck with a slate-grey body, complete with doors, windows, and wheels, positioned on the left side, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A stocky adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in the foreground, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing); keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the cargo utility truck, realized with the materials above. <Subject 5> (the person receiving the object) is a stocky adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing opposite the protagonist, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing), and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a stocky adult fisher with a salt-and-pepper beard and short hair, wearing a rust waterproof jacket, olive waders, and rubber boots, standing in the background, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a stocky adult construction worker with short black hair, wearing an amber high-visibility vest, khaki work trousers, and tan work boots, standing in the background, seen only from behind, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (independently moving adult participant 3) is a stocky adult courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the background, drawn as an anime character in the same cel-shaded anime manner (character design: elegant elongated figures with flowing hair and clothing) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The cargo utility truck is one continuous object throughout and comes to rest about 7 metres from the protagonist. The person receiving the object in the foreground covers independently moving adult participant 2 from about 4.4 s to 4.7 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — cel-shaded anime: flat cel shading, thin clean contour lines, vivid saturated colours and anime facial proportions; a warm amber-and-teal palette, golden-hour side light with long soft shadows, crisp cold air with a hint of frost, detailed line-drawn backgrounds with light washes; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-117-city-combat/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-117-city-combat/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a0c18d66af2582f77439674fc7ebba58331de143
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-117-city-combat/follow/colored-obb/prompt.txt
@@ -0,0 +1,66 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 human proxies standing for <Subject 1> (the protagonist), <Subject 14> (independently moving adult participant 1), <Subject 15> (independently moving adult participant 2), <Subject 16> (independently moving adult participant 3), <Subject 17> (the training opponent moving laterally and returning fire), <Subject 18> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 12> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 10> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 11> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 13> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey finish, positioned in the middle ground. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine, held by the player. A pickup truck with a matte slate-grey body, featuring a complete chassis with doors, windows, and wheels, positioned in the distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the solid proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 10> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the orange flame shape that appears when it is active that starts on the left of the frame; use the target appearance described below.
+<Subject 11> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the orange flame shape that appears when it is active that starts on the left of the frame; use the target appearance described below.
+<Subject 12> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 13> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 14> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 15> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 16> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 17> is the training opponent moving laterally and returning fire; in <Video 1> the human proxy tagged #f73677; use the target appearance described below.
+<Subject 18> is the second training opponent returning fire from the opposite side; in <Video 1> the human proxy tagged #36aff7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 18>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 11> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 13> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 17> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 18> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey finish, positioned in the middle ground. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine, held by the player. A pickup truck with a matte slate-grey body, featuring a complete chassis with doors, windows, and wheels, positioned in the distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult worker in a rugged exosuit with short black hair, wearing a safety-amber exoskeleton frame over a charcoal jumpsuit with leg braces and heavy boots, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the short-lived shot tracer, realized with the materials above. <Subject 11> is the short-lived muzzle flash, realized with the materials above. <Subject 12> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 13> is the short-lived shot tracer, realized with the materials above. <Subject 14> (independently moving adult participant 1) is an adult street vendor with grey-streaked black hair, wearing a khaki apron over a sage tee, tan chinos, and rubber clogs - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (independently moving adult participant 2) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 16> (independently moving adult participant 3) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 17> (the training opponent moving laterally and returning fire) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 18> (the second training opponent returning fire from the opposite side) is an adult paramedic with short auburn hair, wearing an emerald medic jacket with reflective strips, charcoal trousers, and black boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves a short distance along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre; the effect moves about half a metre along its route. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the effect moves a short distance along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 12> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and comes to rest about 14 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and comes to rest about 12 metres from the protagonist. <Subject 12> is one continuous object throughout and comes to rest about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 3.1 s to 3.6 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 3.0 s to 3.6 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 2.1 s to 3.2 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 2.7 s to 3.9 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.2 s to 3.9 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.2 s to 3.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-117-city-combat/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-117-city-combat/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c960b6ae792698b5656028e0c881bad6b5e4bc20
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-117-city-combat/follow/mixed/prompt.txt
@@ -0,0 +1,66 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 14> (independently moving adult participant 1), <Subject 15> (independently moving adult participant 2), <Subject 16> (independently moving adult participant 3), <Subject 17> (the training opponent moving laterally and returning fire), <Subject 18> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 12> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 10> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 11> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 13> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey finish, positioned in the middle ground. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine, held by the player. A pickup truck with a matte slate-grey body, featuring a complete chassis with doors, windows, and wheels, positioned in the distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the black solid obstacle proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 10> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active that starts on the left of the frame; use the target appearance described below.
+<Subject 11> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active that starts on the left of the frame; use the target appearance described below.
+<Subject 12> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 13> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 14> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 15> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 16> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 17> is the training opponent moving laterally and returning fire; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 18> is the second training opponent returning fire from the opposite side; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 18>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 11> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 13> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 17> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 18> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey finish, positioned in the middle ground. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine, held by the player. A pickup truck with a matte slate-grey body, featuring a complete chassis with doors, windows, and wheels, positioned in the distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult worker in a rugged exosuit with short black hair, wearing a safety-amber exoskeleton frame over a charcoal jumpsuit with leg braces and heavy boots, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the short-lived shot tracer, realized with the materials above. <Subject 11> is the short-lived muzzle flash, realized with the materials above. <Subject 12> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 13> is the short-lived shot tracer, realized with the materials above. <Subject 14> (independently moving adult participant 1) is an adult street vendor with grey-streaked black hair, wearing a khaki apron over a sage tee, tan chinos, and rubber clogs - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (independently moving adult participant 2) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 16> (independently moving adult participant 3) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 17> (the training opponent moving laterally and returning fire) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 18> (the second training opponent returning fire from the opposite side) is an adult paramedic with short auburn hair, wearing an emerald medic jacket with reflective strips, charcoal trousers, and black boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves a short distance along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre; the effect moves about half a metre along its route. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the effect moves a short distance along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 12> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and comes to rest about 14 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and comes to rest about 12 metres from the protagonist. <Subject 12> is one continuous object throughout and comes to rest about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 3.1 s to 3.6 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 3.0 s to 3.6 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 2.1 s to 3.2 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 2.7 s to 3.9 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.2 s to 3.9 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.2 s to 3.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-117-city-combat/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-117-city-combat/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f9716ee076d715c7596a1689c731e20536b6530c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-117-city-combat/follow/white/prompt.txt
@@ -0,0 +1,66 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 14> (independently moving adult participant 1), <Subject 15> (independently moving adult participant 2), <Subject 16> (independently moving adult participant 3), <Subject 17> (the training opponent moving laterally and returning fire), <Subject 18> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 12> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 10> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 11> is not in frame at 00:00.000 and enters at about 1.6 s; <Subject 13> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey finish, positioned in the middle ground. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine, held by the player. A pickup truck with a matte slate-grey body, featuring a complete chassis with doors, windows, and wheels, positioned in the distance - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 10> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the grey flame-shaped solid that appears when it is active that starts on the left of the frame; use the target appearance described below.
+<Subject 11> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.6 s; in <Video 1> the grey flame-shaped solid that appears when it is active that starts on the left of the frame; use the target appearance described below.
+<Subject 12> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 13> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 14> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 15> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 16> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 17> is the training opponent moving laterally and returning fire; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 18> is the second training opponent returning fire from the opposite side; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 18>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 11> (enters [Shot 1] at about 1.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 13> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 17> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 18> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey finish, positioned in the middle ground. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine, held by the player. A pickup truck with a matte slate-grey body, featuring a complete chassis with doors, windows, and wheels, positioned in the distance - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult worker in a rugged exosuit with short black hair, wearing a safety-amber exoskeleton frame over a charcoal jumpsuit with leg braces and heavy boots, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the short-lived shot tracer, realized with the materials above. <Subject 11> is the short-lived muzzle flash, realized with the materials above. <Subject 12> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 13> is the short-lived shot tracer, realized with the materials above. <Subject 14> (independently moving adult participant 1) is an adult street vendor with grey-streaked black hair, wearing a khaki apron over a sage tee, tan chinos, and rubber clogs - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (independently moving adult participant 2) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 16> (independently moving adult participant 3) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 17> (the training opponent moving laterally and returning fire) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 18> (the second training opponent returning fire from the opposite side) is an adult paramedic with short auburn hair, wearing an emerald medic jacket with reflective strips, charcoal trousers, and black boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves a short distance along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre; the effect moves about half a metre along its route. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the effect moves a short distance along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 12> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and comes to rest about 14 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and comes to rest about 12 metres from the protagonist. <Subject 12> is one continuous object throughout and comes to rest about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 3.1 s to 3.6 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 3.0 s to 3.6 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 2.1 s to 3.2 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 2.7 s to 3.9 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.2 s to 3.9 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 2.2 s to 3.2 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-118-city-combat/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-118-city-combat/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e0f4565551344fe390b668241c93cec2c84a7cb4
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-118-city-combat/follow/colored-obb/prompt.txt
@@ -0,0 +1,62 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 human proxies standing for <Subject 1> (the protagonist), <Subject 12> (independently moving adult participant 1), <Subject 13> (independently moving adult participant 2), <Subject 14> (independently moving adult participant 3), <Subject 15> (the training opponent moving laterally and returning fire), <Subject 16> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 11> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey surface. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine. A vintage sedan with a faded cream body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 12> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 13> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 14> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 15> is the training opponent moving laterally and returning fire; in <Video 1> the human proxy tagged #f73677; use the target appearance described below.
+<Subject 16> is the second training opponent returning fire from the opposite side; in <Video 1> the human proxy tagged #36aff7; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 16>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey surface. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine. A vintage sedan with a faded cream body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (independently moving adult participant 1) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (independently moving adult participant 2) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 3) is an adult tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (the training opponent moving laterally and returning fire) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the second training opponent returning fire from the opposite side) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves a short distance along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves a short distance along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 13 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and ends the clip about 12 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 3.1 s to 3.4 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 3.2 s to 3.4 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 1.0 s to 1.9 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 4.2 s to 4.7 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 0.9 s to 2.3 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 4.2 s to 5.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 1.2 s to 2.3 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-118-city-combat/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-118-city-combat/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9b20fe0f6613f5d5be9f7e04fda06f09b4d6b817
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-118-city-combat/follow/mixed/prompt.txt
@@ -0,0 +1,62 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 12> (independently moving adult participant 1), <Subject 13> (independently moving adult participant 2), <Subject 14> (independently moving adult participant 3), <Subject 15> (the training opponent moving laterally and returning fire), <Subject 16> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 11> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey surface. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine. A vintage sedan with a faded cream body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 12> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 13> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 14> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 15> is the training opponent moving laterally and returning fire; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 16> is the second training opponent returning fire from the opposite side; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 16>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey surface. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine. A vintage sedan with a faded cream body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (independently moving adult participant 1) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (independently moving adult participant 2) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 3) is an adult tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (the training opponent moving laterally and returning fire) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the second training opponent returning fire from the opposite side) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves a short distance along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves a short distance along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 13 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and ends the clip about 12 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 3.1 s to 3.4 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 3.2 s to 3.4 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 1.0 s to 1.9 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 4.2 s to 4.7 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 0.9 s to 2.3 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 4.2 s to 5.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 1.2 s to 2.3 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-118-city-combat/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-118-city-combat/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9254f3e703dd9c11623a61669bac2aff2b5d8e08
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-118-city-combat/follow/white/prompt.txt
@@ -0,0 +1,62 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 6 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 12> (independently moving adult participant 1), <Subject 13> (independently moving adult participant 2), <Subject 14> (independently moving adult participant 3), <Subject 15> (the training opponent moving laterally and returning fire), <Subject 16> (the second training opponent returning fire from the opposite side) and 6 object proxies standing for <Subject 3> (the passenger car), <Subject 4> (the solid waist-height cover used in both exchanges), <Subject 5> (the solid cover that stops incoming shots), <Subject 6> (the training rifle with stock, receiver, barrel and magazine), <Subject 9> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 8> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 11> is not in frame at 00:00.000 and enters at about 3.2 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey surface. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine. A vintage sedan with a faded cream body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid waist-height cover used in both exchanges; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the solid cover that stops incoming shots; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 7> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 8> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 9> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 3.2 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 12> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 13> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 14> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 15> is the training opponent moving laterally and returning fire; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 16> is the second training opponent returning fire from the opposite side; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 16>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 9> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 3.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 14> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 15> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 16> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A city street with tall buildings, sidewalks, parked cars and crossing traffic. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid waist-height concrete barrier with a rough, weathered grey surface. A training rifle with a matte black finish, featuring a stock, receiver, barrel, and magazine. A vintage sedan with a faded cream body, complete with doors, windows, and wheels, positioned on the street - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, holding a training rifle, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the passenger car, realized with the materials above. <Subject 4> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 5> is the solid cover that stops incoming shots, realized with the materials above. <Subject 6> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 7> is the short-lived shot tracer, realized with the materials above. <Subject 8> is the short-lived muzzle flash, realized with the materials above. <Subject 9> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (independently moving adult participant 1) is an adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (independently moving adult participant 2) is an adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 14> (independently moving adult participant 3) is an adult tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 15> (the training opponent moving laterally and returning fire) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 16> (the second training opponent returning fire from the opposite side) is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, holding a training rifle - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about a metre along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves a short distance along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), straightens the legs; independently moving adult participant 1 moves a short distance along its route; the vehicle moves about half a metre along its route; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves a short distance along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist holds <Subject 6> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 9> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 13 metres from the protagonist. <Subject 6> is one continuous object throughout and comes to rest about half a metre from the protagonist. <Subject 9> is one continuous object throughout and ends the clip about 12 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 14 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 2 from about 3.1 s to 3.4 s; the short-lived shot tracer in the foreground covers independently moving adult participant 2 from about 3.2 s to 3.4 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the short-lived shot tracer in the foreground covers independently moving adult participant 3 from about 1.4 s to 1.8 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 1.0 s to 1.9 s; the protagonist in the foreground covers the training opponent moving laterally and returning fire from about 4.2 s to 4.7 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 0.9 s to 2.3 s; the solid cover that stops incoming shots in the foreground covers the training opponent moving laterally and returning fire from about 4.2 s to 5.0 s; the short-lived shot tracer in the foreground covers the training opponent moving laterally and returning fire from about 1.2 s to 2.3 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fa4df69ce2d554bbdd68fbdd9619d5c7f46bfbbd
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the cargo utility truck), <Subject 4> (the same distinctive three-tier landmark seen before departure and after return); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A distinctive three-tier landmark sculpture constructed from brushed steel and dark iron, featuring geometric angles and a weathered finish, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds). A six-wheel rover with a matte slate-grey body, heavy-duty tires, and reinforced plating, positioned on the wet deck, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the cargo utility truck; in <Video 1> the object proxy tagged #f79e36 that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the same distinctive three-tier landmark seen before departure and after return; in <Video 1> the solid proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking away from a landmark, scanning the surroundings and returning to it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A distinctive three-tier landmark sculpture constructed from brushed steel and dark iron, featuring geometric angles and a weathered finish, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds). A six-wheel rover with a matte slate-grey body, heavy-duty tires, and reinforced plating, positioned on the wet deck, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same heavy impasto oil manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the cargo utility truck, realized with the materials above. <Subject 4> is the same distinctive three-tier landmark seen before departure and after return, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, seen only from behind, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is an adult diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels and fins, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.633, the protagonist stands and looks at the nearby landmark. From 00:00.633 to 00:02.133, the protagonist walks away from the landmark; the protagonist moves about 4 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about one and a half metres along its route. From 00:02.133 to 00:02.800, the protagonist pauses at a distance and scans the surroundings; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:02.800 to 00:04.433, the protagonist walks back toward the landmark; the protagonist moves about 4 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about 2 metres along its route. From 00:04.433 to 00:05.000, the protagonist arrives back at the landmark and looks at it again; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The cargo utility truck is one continuous object throughout and ends the clip about 14 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6a7cce094bb3b6fd7f362435975eeca1cdab0fb1
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the cargo utility truck), <Subject 4> (the same distinctive three-tier landmark seen before departure and after return); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A distinctive three-tier landmark sculpture constructed from brushed steel and dark iron, featuring geometric angles and a weathered finish, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds). A six-wheel rover with a matte slate-grey body, heavy-duty tires, and reinforced plating, positioned on the wet deck, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the cargo utility truck; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the same distinctive three-tier landmark seen before departure and after return; in <Video 1> the black solid sculpture proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking away from a landmark, scanning the surroundings and returning to it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A distinctive three-tier landmark sculpture constructed from brushed steel and dark iron, featuring geometric angles and a weathered finish, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds). A six-wheel rover with a matte slate-grey body, heavy-duty tires, and reinforced plating, positioned on the wet deck, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same heavy impasto oil manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the cargo utility truck, realized with the materials above. <Subject 4> is the same distinctive three-tier landmark seen before departure and after return, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, seen only from behind, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is an adult diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels and fins, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.633, the protagonist stands and looks at the nearby landmark. From 00:00.633 to 00:02.133, the protagonist walks away from the landmark; the protagonist moves about 4 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about one and a half metres along its route. From 00:02.133 to 00:02.800, the protagonist pauses at a distance and scans the surroundings; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:02.800 to 00:04.433, the protagonist walks back toward the landmark; the protagonist moves about 4 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about 2 metres along its route. From 00:04.433 to 00:05.000, the protagonist arrives back at the landmark and looks at it again; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The cargo utility truck is one continuous object throughout and ends the clip about 14 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..78ef6c5963ee71103e4c0b7cacf2f911abe93732
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-120-city-revisit/first-person/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the cargo utility truck), <Subject 4> (the same distinctive three-tier landmark seen before departure and after return); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A distinctive three-tier landmark sculpture constructed from brushed steel and dark iron, featuring geometric angles and a weathered finish, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds). A six-wheel rover with a matte slate-grey body, heavy-duty tires, and reinforced plating, positioned on the wet deck, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the cargo utility truck; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the same distinctive three-tier landmark seen before departure and after return; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking away from a landmark, scanning the surroundings and returning to it, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A working port with stacked shipping containers, a quay, cranes and cargo vehicles. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A distinctive three-tier landmark sculpture constructed from brushed steel and dark iron, featuring geometric angles and a weathered finish, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds). A six-wheel rover with a matte slate-grey body, heavy-duty tires, and reinforced plating, positioned on the wet deck, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: an adult protagonist in contemporary clothing painted in the same heavy impasto oil manner, with painted features, hair and outfit; keep this identity and outfit consistent throughout the shot. <Subject 3> is the cargo utility truck, realized with the materials above. <Subject 4> is the same distinctive three-tier landmark seen before departure and after return, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, seen only from behind, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is an adult diver with dark hair under a hood, wearing a charcoal wetsuit with slate panels and fins, painted in the same heavy impasto oil manner (character design: figures with warm glazed skin and bold cloth folds) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.633, the protagonist stands and looks at the nearby landmark. From 00:00.633 to 00:02.133, the protagonist walks away from the landmark; the protagonist moves about 4 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about one and a half metres along its route. From 00:02.133 to 00:02.800, the protagonist pauses at a distance and scans the surroundings; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. From 00:02.800 to 00:04.433, the protagonist walks back toward the landmark; the protagonist moves about 4 metres along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the vehicle moves about 2 metres along its route. From 00:04.433 to 00:05.000, the protagonist arrives back at the landmark and looks at it again; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the vehicle moves about half a metre along its route. The cargo utility truck is one continuous object throughout and ends the clip about 14 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, heavy impasto oil look: paint piled in thick sculpted ridges that catch the light, coarse canvas showing at the edges, bold direct colour instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a cool blue-grey palette with muted warm accents, flat overcast daylight, clear dry air, broad loaded brush strokes, background detail kept as thin sheer paint; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7d7d8dd2db200f35dc30fefac549cb514aa98434
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 human proxies standing for <Subject 6> (independently moving adult participant 1), <Subject 7> (the person pushing the wheeled handcart) and 3 object proxies standing for <Subject 3> (the full-height sliding cabinet door fitted to the wall opening), <Subject 4> (the parcel or cargo crate), <Subject 5> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned to the left, its surface showing signs of use with a matte finish - at the start of the clip its front toward the camera. A full-height sliding cabinet door made of smooth, matte-finished wood, fitted seamlessly into the wall opening, reflecting the dim light subtly. A small, rectangular cargo crate made of weathered cardboard, supported by the cabinet shelf in the middle distance, its surface slightly worn and textured. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the full-height sliding cabinet door fitted to the wall opening; in <Video 1> the solid proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is the parcel or cargo crate; in <Video 1> the solid proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the object proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is opening a cabinet door and taking out a parcel, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned to the left, its surface showing signs of use with a matte finish - at the start of the clip its front toward the camera. A full-height sliding cabinet door made of smooth, matte-finished wood, fitted seamlessly into the wall opening, reflecting the dim light subtly. A small, rectangular cargo crate made of weathered cardboard, supported by the cabinet shelf in the middle distance, its surface slightly worn and textured. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the full-height sliding cabinet door fitted to the wall opening, realized with the materials above. <Subject 4> is the parcel or cargo crate, realized with the materials above. <Subject 5> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the upper center of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 1.2 s; from then on it is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, standing in the upper left of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the cabinet; the protagonist moves about half a metre along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; the cart moves about a metre along its route. From 00:00.967 to 00:02.067, the protagonist swings the cabinet door open; the door moves about 3 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.067 to 00:03.367, the protagonist reaches into the open cabinet and takes hold of the parcel; the protagonist moves about a metre along its route, raises the hands to chest height; the crate moves a short distance along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:03.367 to 00:05.000, the protagonist takes the parcel out of the cabinet and holds it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist opens the full-height sliding cabinet door fitted to the wall opening from 00:00.000 to 00:02.067; the protagonist passed through the full-height sliding cabinet door fitted to the wall opening from 00:02.067 to 00:05.000; the protagonist holds the parcel or cargo crate from 00:02.133 to 00:05.000. The full-height sliding cabinet door fitted to the wall opening is one continuous object throughout and comes to rest about 4 metres from the protagonist. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The full-height sliding cabinet door fitted to the wall opening in the foreground covers independently moving adult participant 2 from about 0.0 s to 1.2 s; the wheeled handcart in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.2 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5869356d6897a9342ba3231e7346118d64e07abf
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 blue mannequin proxies standing for <Subject 6> (independently moving adult participant 1), <Subject 7> (the person pushing the wheeled handcart) and 3 object proxies standing for <Subject 3> (the full-height sliding cabinet door fitted to the wall opening), <Subject 4> (the parcel or cargo crate), <Subject 5> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned to the left, its surface showing signs of use with a matte finish - at the start of the clip its front toward the camera. A full-height sliding cabinet door made of smooth, matte-finished wood, fitted seamlessly into the wall opening, reflecting the dim light subtly. A small, rectangular cargo crate made of weathered cardboard, supported by the cabinet shelf in the middle distance, its surface slightly worn and textured. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the full-height sliding cabinet door fitted to the wall opening; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is the parcel or cargo crate; in <Video 1> the red box proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the black solid cart proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is opening a cabinet door and taking out a parcel, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned to the left, its surface showing signs of use with a matte finish - at the start of the clip its front toward the camera. A full-height sliding cabinet door made of smooth, matte-finished wood, fitted seamlessly into the wall opening, reflecting the dim light subtly. A small, rectangular cargo crate made of weathered cardboard, supported by the cabinet shelf in the middle distance, its surface slightly worn and textured. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the full-height sliding cabinet door fitted to the wall opening, realized with the materials above. <Subject 4> is the parcel or cargo crate, realized with the materials above. <Subject 5> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the upper center of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 1.2 s; from then on it is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, standing in the upper left of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the cabinet; the protagonist moves about half a metre along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; the cart moves about a metre along its route. From 00:00.967 to 00:02.067, the protagonist swings the cabinet door open; the door moves about 3 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.067 to 00:03.367, the protagonist reaches into the open cabinet and takes hold of the parcel; the protagonist moves about a metre along its route, raises the hands to chest height; the crate moves a short distance along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:03.367 to 00:05.000, the protagonist takes the parcel out of the cabinet and holds it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist opens the full-height sliding cabinet door fitted to the wall opening from 00:00.000 to 00:02.067; the protagonist passed through the full-height sliding cabinet door fitted to the wall opening from 00:02.067 to 00:05.000; the protagonist holds the parcel or cargo crate from 00:02.133 to 00:05.000. The full-height sliding cabinet door fitted to the wall opening is one continuous object throughout and comes to rest about 4 metres from the protagonist. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The full-height sliding cabinet door fitted to the wall opening in the foreground covers independently moving adult participant 2 from about 0.0 s to 1.2 s; the wheeled handcart in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.2 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5c2f856f5143b2c7025194635a6148d67181452a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-121-indoor-cabinet/first-person/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 grey block-figure proxies standing for <Subject 6> (independently moving adult participant 1), <Subject 7> (the person pushing the wheeled handcart) and 3 object proxies standing for <Subject 3> (the full-height sliding cabinet door fitted to the wall opening), <Subject 4> (the parcel or cargo crate), <Subject 5> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 5> is not in frame at 00:00.000 and enters at about 1.2 s; <Subject 7> is not in frame at 00:00.000 and enters at about 1.2 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned to the left, its surface showing signs of use with a matte finish - at the start of the clip its front toward the camera. A full-height sliding cabinet door made of smooth, matte-finished wood, fitted seamlessly into the wall opening, reflecting the dim light subtly. A small, rectangular cargo crate made of weathered cardboard, supported by the cabinet shelf in the middle distance, its surface slightly worn and textured. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the full-height sliding cabinet door fitted to the wall opening; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 1.2 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is opening a cabinet door and taking out a parcel, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (enters [Shot 1] at about 1.2 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (enters [Shot 1] at about 1.2 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden platform, positioned to the left, its surface showing signs of use with a matte finish - at the start of the clip its front toward the camera. A full-height sliding cabinet door made of smooth, matte-finished wood, fitted seamlessly into the wall opening, reflecting the dim light subtly. A small, rectangular cargo crate made of weathered cardboard, supported by the cabinet shelf in the middle distance, its surface slightly worn and textured. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the full-height sliding cabinet door fitted to the wall opening, realized with the materials above. <Subject 4> is the parcel or cargo crate, realized with the materials above. <Subject 5> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 6> (independently moving adult participant 1) is an adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the upper center of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 1.2 s; from then on it is an adult mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, steel-toe boots, standing in the upper left of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist walks up to the cabinet; the protagonist moves about half a metre along its route; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; the cart moves about a metre along its route. From 00:00.967 to 00:02.067, the protagonist swings the cabinet door open; the door moves about 3 metres along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.067 to 00:03.367, the protagonist reaches into the open cabinet and takes hold of the parcel; the protagonist moves about a metre along its route, raises the hands to chest height; the crate moves a short distance along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:03.367 to 00:05.000, the protagonist takes the parcel out of the cabinet and holds it; independently moving adult participant 1 moves about one and a half metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist opens the full-height sliding cabinet door fitted to the wall opening from 00:00.000 to 00:02.067; the protagonist passed through the full-height sliding cabinet door fitted to the wall opening from 00:02.067 to 00:05.000; the protagonist holds the parcel or cargo crate from 00:02.133 to 00:05.000. The full-height sliding cabinet door fitted to the wall opening is one continuous object throughout and comes to rest about 4 metres from the protagonist. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The full-height sliding cabinet door fitted to the wall opening in the foreground covers independently moving adult participant 2 from about 0.0 s to 1.2 s; the wheeled handcart in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.2 s. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-122-indoor-place/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-122-indoor-place/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fca05252979030f2dcb6cfac1286040775d9e183
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-122-indoor-place/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of rusted iron and dark wood, positioned in the background to the left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip its front toward the camera. A sturdy cargo crate made of dark, weathered wood with metal corner reinforcements, supported by the pickup stand in the middle ground, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #6636f7; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of rusted iron and dark wood, positioned in the background to the left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip its front toward the camera. A sturdy cargo crate made of dark, weathered wood with metal corner reinforcements, supported by the pickup stand in the middle ground, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines). <Subject 1>, the protagonist, has the following target appearance: A mechanic with hair tied back, wearing work overalls and steel-toe boots, standing centrally in the foreground, seen only from behind, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a windbreaker, slate cargo shorts, and ivory sneakers, standing to the middle left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an office worker with neat short hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing to the middle right, seen only from behind, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing further back in the middle left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 2.4 s to 2.7 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 2.9 s to 3.2 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 1.9 s; the room-lintel in the foreground covers independently moving adult participant 2 from about 2.2 s to 2.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.9 s to 2.9 s; the room-lintel in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-122-indoor-place/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-122-indoor-place/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..811032d424178bd0b24285c1195fb72c8dcd538a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-122-indoor-place/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of rusted iron and dark wood, positioned in the background to the left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip its front toward the camera. A sturdy cargo crate made of dark, weathered wood with metal corner reinforcements, supported by the pickup stand in the middle ground, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines). In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of rusted iron and dark wood, positioned in the background to the left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip its front toward the camera. A sturdy cargo crate made of dark, weathered wood with metal corner reinforcements, supported by the pickup stand in the middle ground, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines). <Subject 1>, the protagonist, has the following target appearance: A mechanic with hair tied back, wearing work overalls and steel-toe boots, standing centrally in the foreground, seen only from behind, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a windbreaker, slate cargo shorts, and ivory sneakers, standing to the middle left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an office worker with neat short hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing to the middle right, seen only from behind, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing further back in the middle left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 2.4 s to 2.7 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 2.9 s to 3.2 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 1.9 s; the room-lintel in the foreground covers independently moving adult participant 2 from about 2.2 s to 2.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.9 s to 2.9 s; the room-lintel in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-122-indoor-place/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-122-indoor-place/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..006f263db935a0e12c504ba1fe2de71d42942425
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-122-indoor-place/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of rusted iron and dark wood, positioned in the background to the left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip its front toward the camera. A sturdy cargo crate made of dark, weathered wood with metal corner reinforcements, supported by the pickup stand in the middle ground, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines). In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it forward and placing it on a table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart constructed of rusted iron and dark wood, positioned in the background to the left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip its front toward the camera. A sturdy cargo crate made of dark, weathered wood with metal corner reinforcements, supported by the pickup stand in the middle ground, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines). <Subject 1>, the protagonist, has the following target appearance: A mechanic with hair tied back, wearing work overalls and steel-toe boots, standing centrally in the foreground, seen only from behind, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a windbreaker, slate cargo shorts, and ivory sneakers, standing to the middle left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is an office worker with neat short hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing to the middle right, seen only from behind, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is a student with shoulder-length hair, wearing a hoodie, denim jeans, and ivory canvas sneakers, standing further back in the middle left, drawn in the same fine crosshatched etching-style drawing manner (character design: figures built from hatching with almost no outlines) - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel forward; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the table and the protagonist lets go of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 2.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 2.4 s to 2.7 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 2.9 s to 3.2 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.9 s to 1.9 s; the room-lintel in the foreground covers independently moving adult participant 2 from about 2.2 s to 2.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.9 s to 2.9 s; the room-lintel in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.5 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 2.9 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, etching-style drawing look: extremely fine black crosshatched lines building every tone, crisp engraved edges, pure white paper, no colour, drawn over the full 3D world so every recorded shape stays readable; a monochrome tonal palette, daylight through tall windows with soft shadows, a slight interior haze, background fading out to blank paper at the edges; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..136bad123dba75fcb18544073a050485d6d7af87
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden shelves, positioned on the far left side of the room, partially cut off by the frame edge - at the start of the clip its front toward the camera. A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist near the center of the room. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the object proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden shelves, positioned on the far left side of the room, partially cut off by the frame edge - at the start of the clip its front toward the camera. A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist near the center of the room. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the room, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing near the center of the room - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing on the far left side of the room - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a chef with dark hair, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs, standing in the middle left of the room, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, and steel-toe boots, standing on the far left side of the room, partially cut off by the frame edge - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..97acd4d51b728cf599820bb03cb94d3a08b59678
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden shelves, positioned on the far left side of the room, partially cut off by the frame edge - at the start of the clip its front toward the camera. A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist near the center of the room. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden shelves, positioned on the far left side of the room, partially cut off by the frame edge - at the start of the clip its front toward the camera. A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist near the center of the room. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the room, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing near the center of the room - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing on the far left side of the room - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a chef with dark hair, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs, standing in the middle left of the room, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, and steel-toe boots, standing on the far left side of the room, partially cut off by the frame edge - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f561f402c940d0fecf09f5e4eaf8065699b22ac9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-124-indoor-handoff/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden shelves, positioned on the far left side of the room, partially cut off by the frame edge - at the start of the clip its front toward the camera. A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist near the center of the room. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a metal frame and wooden shelves, positioned on the far left side of the room, partially cut off by the frame edge - at the start of the clip its front toward the camera. A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist near the center of the room. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the room, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is an adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing near the center of the room - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing on the far left side of the room - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a chef with dark hair, wearing an ivory double-breasted chef jacket, houndstooth trousers, and black kitchen clogs, standing in the middle left of the room, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee, and steel-toe boots, standing on the far left side of the room, partially cut off by the frame edge - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about 2 metres along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 4 metres from the protagonist. The wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..55c923c7f00d626dcc9fd10e83e4ef9c3a589c29
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 3> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 4> is not in frame at 00:00.000 and enters at about 1.0 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the object proxy tagged #f79e36 that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1: not in frame at 00:00.000; it enters the frame at about 1.0 s; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 4> (enters [Shot 1] at about 1.0 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: A realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is not in frame at 00:00.000 and enters the frame at about 1.0 s; from then on it is a realistic adult site engineer with short wavy dark hair, wearing a cream polo under a navy fleece, slate trousers, and brown boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it keeps one consistent plausible appearance in the target style. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The wheeled handcart is one continuous object throughout and ends the clip about 6 metres from the protagonist. Independently moving adult participant 1 in the foreground covers the protagonist from about 2.9 s to 3.5 s; the solid wall or architectural structure in the foreground covers the protagonist from about 4.1 s to 5.0 s; the protagonist in the foreground covers independently moving adult participant 2 from about 3.6 s to 3.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.6 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ca742ad0bee8ff7c3a445439db77252e2a46e944
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 3> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 4> is not in frame at 00:00.000 and enters at about 1.0 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the black solid cart proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1: not in frame at 00:00.000; it enters the frame at about 1.0 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 4> (enters [Shot 1] at about 1.0 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: A realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is not in frame at 00:00.000 and enters the frame at about 1.0 s; from then on it is a realistic adult site engineer with short wavy dark hair, wearing a cream polo under a navy fleece, slate trousers, and brown boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it keeps one consistent plausible appearance in the target style. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The wheeled handcart is one continuous object throughout and ends the clip about 6 metres from the protagonist. Independently moving adult participant 1 in the foreground covers the protagonist from about 2.9 s to 3.5 s; the solid wall or architectural structure in the foreground covers the protagonist from about 4.1 s to 5.0 s; the protagonist in the foreground covers independently moving adult participant 2 from about 3.6 s to 3.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.6 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9e327deff41ed0356c8d6b755516e32c775c91a2
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-125-indoor-walk/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 3> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 4> is not in frame at 00:00.000 and enters at about 1.0 s; <Subject 6> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1: not in frame at 00:00.000; it enters the frame at about 1.0 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking forward around a bend of the trail, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 4> (enters [Shot 1] at about 1.0 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: A realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is not in frame at 00:00.000 and enters the frame at about 1.0 s; from then on it is a realistic adult site engineer with short wavy dark hair, wearing a cream polo under a navy fleece, slate trousers, and brown boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it keeps one consistent plausible appearance in the target style. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward along the trail and around its bend; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 4 metres along its route; independently moving adult participant 3 moves about 4 metres along its route; the cart moves about 4 metres along its route. From 00:04.767 to 00:05.000, the protagonist continues walking slowly. The wheeled handcart is one continuous object throughout and ends the clip about 6 metres from the protagonist. Independently moving adult participant 1 in the foreground covers the protagonist from about 2.9 s to 3.5 s; the solid wall or architectural structure in the foreground covers the protagonist from about 4.1 s to 5.0 s; the protagonist in the foreground covers independently moving adult participant 2 from about 3.6 s to 3.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 0.0 s to 0.5 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 2.3 s to 2.6 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..06afbea5547f9d04f1d37eb9c90de407720bab61
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, positioned in the middle ground - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the object proxy tagged #f79e36; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> advancing through independently moving pedestrians and a handcart, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, positioned in the middle ground - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, standing further back in the center - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward through the moving pedestrians while the handcart is pushed along nearby; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the others keep moving. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.6 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.2 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 2.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.3 s to 3.0 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2649a6a37cd3efb2fd0342f432edf51472b4ff29
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, positioned in the middle ground - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> advancing through independently moving pedestrians and a handcart, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, positioned in the middle ground - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, standing further back in the center - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward through the moving pedestrians while the handcart is pushed along nearby; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the others keep moving. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.6 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.2 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 2.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.3 s to 3.0 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f6b91ff15b4477b50af1bc8043e31f8ab72838a1
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-128-indoor-crowd/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, positioned in the middle ground - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> advancing through independently moving pedestrians and a handcart, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed office with corridors, sliding doors, desks and stairs between floors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A wheeled handcart with a matte grey metal frame and black rubber wheels, positioned in the middle ground - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult office worker with neat short chestnut hair, wearing an ivory shirt with a plum tie, charcoal suit trousers, and polished oxford shoes, standing to the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult nurse with dark hair in a bun, wearing a sage scrubs top, sage scrubs trousers, and ivory clogs, standing further back in the center - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward through the moving pedestrians while the handcart is pushed along nearby; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the others keep moving. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.0 s to 0.6 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.2 s to 1.5 s; the protagonist in the foreground covers independently moving adult participant 3 from about 1.4 s to 2.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.3 s to 3.0 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c149c0e058ba98abcdf883b3a171769892a85e3d
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/colored-obb/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 8> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the object proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is a young adult student with shoulder-length ash-brown hair, dressed in a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a security guard with a shaved head, clad in a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it keeps one consistent plausible appearance in the target style. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 3 metres from the protagonist. The storage rack with varied stock and empty spaces in the foreground covers independently moving adult participant 3 from about 2.6 s to 3.3 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 3.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e81e218aa76af1ddbe1033165343ad813a292e21
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/mixed/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 8> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the black solid cart proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 8> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is a young adult student with shoulder-length ash-brown hair, dressed in a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a security guard with a shaved head, clad in a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it keeps one consistent plausible appearance in the target style. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 3 metres from the protagonist. The storage rack with varied stock and empty spaces in the foreground covers independently moving adult participant 3 from about 2.6 s to 3.3 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 3.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5e222a4824d036f4a73bb225e6c573b95f0f5096
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-129-indoor-handoff/follow/white/prompt.txt
@@ -0,0 +1,46 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 5 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (the person receiving the object), <Subject 6> (independently moving adult participant 1), <Subject 7> (independently moving adult participant 2), <Subject 8> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 2.4 s; <Subject 8> is not in frame at 00:00.000 and enters at about 2.4 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the person receiving the object; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 8> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 2.4 s; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> offering a parcel to another person and releasing it to the receiver, involving <Subject 3> to <Subject 8>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 2.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 8> (enters [Shot 1] at about 2.4 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A rectangular cargo crate made of sturdy cardboard with a matte finish, held by the protagonist. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 8> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (the person receiving the object) is a young adult student with shoulder-length ash-brown hair, dressed in a maroon hoodie, indigo denim jeans, and ivory canvas sneakers - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 1) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 2) is a security guard with a shaved head, clad in a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 8> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 2.4 s; from then on it keeps one consistent plausible appearance in the target style. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.167, the protagonist carries the parcel toward the waiting companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:01.167 to 00:02.467, the protagonist extends the parcel toward the companion, who reaches out for it; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the cart moves about a metre along its route. From 00:02.467 to 00:03.267, both hold the parcel briefly as the grip passes from the protagonist to the companion; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about a metre along its route. From 00:03.267 to 00:05.000, the companion holds the parcel and the protagonist lowers the empty hands; the protagonist moves about half a metre along its route; the person receiving the object moves about half a metre along its route; independently moving adult participant 1 moves about a metre along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about one and a half metres along its route; the cart moves about one and a half metres along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.000 to 00:03.267; the person receiving the object holds the parcel or cargo crate from 00:03.267 to 00:05.000. The parcel or cargo crate is one continuous object throughout and comes to rest about one and a half metres from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 3 metres from the protagonist. The storage rack with varied stock and empty spaces in the foreground covers independently moving adult participant 3 from about 2.6 s to 3.3 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 3.3 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4a2aa6c86e43c681586fabea45dab4224df99539
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/colored-obb/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A standard wheeled handcart with a grey metal wire basket and black rubber wheels, positioned in the middle left of the frame - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the object proxy tagged #f79e36; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the human proxy tagged #a7f736; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f736df; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> advancing through independently moving pedestrians and a handcart, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A standard wheeled handcart with a grey metal wire basket and black rubber wheels, positioned in the middle left of the frame - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the upper centre of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle centre of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward through the moving pedestrians while the handcart is pushed along nearby; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the others keep moving. The wheeled handcart is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 1.1 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 1.1 s to 3.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.5 s to 2.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.9 s; the storage rack with varied stock and empty spaces in the foreground covers independently moving adult participant 3 from about 3.3 s to 4.1 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2fde7e74b8abe12fff1634567ffb7d7dbf360b2f
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/mixed/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A standard wheeled handcart with a grey metal wire basket and black rubber wheels, positioned in the middle left of the frame - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> advancing through independently moving pedestrians and a handcart, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A standard wheeled handcart with a grey metal wire basket and black rubber wheels, positioned in the middle left of the frame - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the upper centre of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle centre of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward through the moving pedestrians while the handcart is pushed along nearby; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the others keep moving. The wheeled handcart is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 1.1 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 1.1 s to 3.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.5 s to 2.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.9 s; the storage rack with varied stock and empty spaces in the foreground covers independently moving adult participant 3 from about 3.3 s to 4.1 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..00680aec3e88f200470ac9af1827aa4b9cb71627
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-131-indoor-crowd/follow/white/prompt.txt
@@ -0,0 +1,42 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 4> (independently moving adult participant 1), <Subject 5> (independently moving adult participant 2), <Subject 6> (the person pushing the wheeled handcart) and 1 object proxy standing for <Subject 3> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A standard wheeled handcart with a grey metal wire basket and black rubber wheels, positioned in the middle left of the frame - at the start of the clip its front toward the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 5> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 6> is the person pushing the wheeled handcart; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> advancing through independently moving pedestrians and a handcart, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully enclosed shop with stocked shelves, shopping carts, a counter and automatic doors. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A standard wheeled handcart with a grey metal wire basket and black rubber wheels, positioned in the middle left of the frame - at the start of the clip its front toward the camera. <Subject 1>, the protagonist, has the following target appearance: A realistic adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the wheeled handcart, pushed by <Subject 6> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 4> (independently moving adult participant 1) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the upper centre of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 5> (independently moving adult participant 2) is a realistic adult tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing in the middle centre of the frame, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (the person pushing the wheeled handcart) is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing in the middle centre of the frame - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks forward through the moving pedestrians while the handcart is pushed along nearby; the protagonist moves about 9 metres along its route; independently moving adult participant 1 moves about 4 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; the cart moves about 3 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop and stands still while the others keep moving. The wheeled handcart is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers independently moving adult participant 1 from about 1.1 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 1.1 s to 3.1 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.5 s to 2.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.9 s; the storage rack with varied stock and empty spaces in the foreground covers independently moving adult participant 3 from about 3.3 s to 4.1 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.5 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7c84d61ea8536aa3d966acdb5596126e730f8c75
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.5 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and rubber wheels, positioned in the background aisle. A sturdy wooden cargo crate with metal corner reinforcements, supported by the pickup stand in the middle ground. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #e7f736 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #ce36f7; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.5 s; in <Video 1> the human proxy tagged #36f796; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it up steps and placing it on a raised table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (enters [Shot 1] at about 0.5 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and rubber wheels, positioned in the background aisle. A sturdy wooden cargo crate with metal corner reinforcements, supported by the pickup stand in the middle ground. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 0.5 s; from then on it is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel up the steps; the protagonist moves about 9 metres along its route, rises about a metre; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the raised table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the raised table and the protagonist lets go of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The upper-platform in the foreground covers the protagonist from about 3.8 s to 5.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.7 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.4 s to 1.7 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.2 s to 1.7 s; the upper-platform in the foreground covers independently moving adult participant 3 from about 3.1 s to 4.6 s; the platform-rail-1 in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s; the platform-rail-1 in the foreground covers independently moving adult participant 3 from about 2.9 s to 4.6 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.9 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 2.8 s to 3.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..26f5c604465b51f189f44ec5d73d0900b8bf8189
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.5 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and rubber wheels, positioned in the background aisle. A sturdy wooden cargo crate with metal corner reinforcements, supported by the pickup stand in the middle ground. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the black solid cart proxy shape; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.5 s; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it up steps and placing it on a raised table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (enters [Shot 1] at about 0.5 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and rubber wheels, positioned in the background aisle. A sturdy wooden cargo crate with metal corner reinforcements, supported by the pickup stand in the middle ground. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 0.5 s; from then on it is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel up the steps; the protagonist moves about 9 metres along its route, rises about a metre; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the raised table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the raised table and the protagonist lets go of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The upper-platform in the foreground covers the protagonist from about 3.8 s to 5.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.7 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.4 s to 1.7 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.2 s to 1.7 s; the upper-platform in the foreground covers independently moving adult participant 3 from about 3.1 s to 4.6 s; the platform-rail-1 in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s; the platform-rail-1 in the foreground covers independently moving adult participant 3 from about 2.9 s to 4.6 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.9 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 2.8 s to 3.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..670f8962efcd533b35584a5536aa616f58289916
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-133-indoor-carryStairs/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (the person pushing the wheeled handcart) and 2 object proxies standing for <Subject 3> (the parcel or cargo crate), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object. <Subject 7> is not in frame at 00:00.000 and enters at about 0.5 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and rubber wheels, positioned in the background aisle. A sturdy wooden cargo crate with metal corner reinforcements, supported by the pickup stand in the middle ground. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the parcel or cargo crate; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the wheeled handcart; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 7> is the person pushing the wheeled handcart: not in frame at 00:00.000; it enters the frame at about 0.5 s; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> lifting a parcel, carrying it up steps and placing it on a raised table, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (enters [Shot 1] at about 0.5 s): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target appearance described below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A utilitarian wheeled handcart with a metal frame and rubber wheels, positioned in the background aisle. A sturdy wooden cargo crate with metal corner reinforcements, supported by the pickup stand in the middle ground. <Subject 1>, the protagonist, has the following target appearance: A realistic adult mechanic with auburn hair tied back, wearing navy work overalls and steel-toe boots, standing in the foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the parcel or cargo crate, realized with the materials above. <Subject 4> is the wheeled handcart, pushed by <Subject 7> who walks directly behind it with both hands on its handle - the cart never moves on its own and the two move together as one unit, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a realistic adult security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a realistic adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 7> (the person pushing the wheeled handcart) is not in frame at 00:00.000 and enters the frame at about 0.5 s; from then on it is a realistic adult bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.733, the protagonist bends toward the parcel and takes hold of it; independently moving adult participant 1 moves a short distance along its route; independently moving adult participant 2 moves a short distance along its route. From 00:00.733 to 00:01.467, the protagonist lifts the parcel off its stand; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:01.467 to 00:03.967, the protagonist carries the parcel up the steps; the protagonist moves about 9 metres along its route, rises about a metre; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about one and a half metres along its route; independently moving adult participant 3 moves about 2 metres along its route; the cart moves about 2 metres along its route. From 00:03.967 to 00:04.567, the protagonist lowers the parcel onto the raised table; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about half a metre along its route; the cart moves about half a metre along its route. From 00:04.567 to 00:05.000, the parcel rests on the raised table and the protagonist lets go of it; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves a short distance along its route; independently moving adult participant 3 moves a short distance along its route; the cart moves a short distance along its route. Recorded relations: the protagonist holds the parcel or cargo crate from 00:00.733 to 00:04.567. The parcel or cargo crate is one continuous object throughout and comes to rest about half a metre from the protagonist. The wheeled handcart is one continuous object throughout and ends the clip about 5 metres from the protagonist. The upper-platform in the foreground covers the protagonist from about 3.8 s to 5.0 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 1 from about 0.7 s to 1.9 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 2 from about 1.4 s to 1.7 s; the solid wall or architectural structure in the foreground covers independently moving adult participant 3 from about 1.2 s to 1.7 s; the upper-platform in the foreground covers independently moving adult participant 3 from about 3.1 s to 4.6 s; the platform-rail-1 in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.0 s; the platform-rail-1 in the foreground covers independently moving adult participant 3 from about 2.9 s to 4.6 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 0.0 s to 1.9 s; the wheeled handcart in the foreground covers independently moving adult participant 3 from about 2.8 s to 3.4 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d753ced3cb5978303391cd5a06a4bce3f7e72097
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 human proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek white body and tinted windows, positioned in the background, its form softened by the focus - at the start of the clip its front away from the camera. An open-top buggy with a matte slate body and exposed mechanical details, positioned in the lower right foreground, partially cut off by the frame edge - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the object proxy tagged #36aff7 that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #6636f7 that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #3ef736 that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the human proxy tagged #f73677 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking up the entry steps into a stationary passenger bus, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek white body and tinted windows, positioned in the background, its form softened by the focus - at the start of the clip its front away from the camera. An open-top buggy with a matte slate body and exposed mechanical details, positioned in the lower right foreground, partially cut off by the frame edge - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the center foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the left of the protagonist, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing on the far left, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.667, the protagonist walks up the entry steps of the stationary bus; the protagonist moves about 2 metres along its route, rises a short distance; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about 2 metres along its route. From 00:01.667 to 00:05.000, the protagonist moves along the aisle inside the bus; the protagonist moves about 4 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 3 moves about 2 metres along its route; the vehicle moves about 4 metres along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:01.667 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 3 metres from the protagonist. The enclosed passenger bus in the foreground covers the protagonist from about 1.2 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 2 from about 3.8 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.9 s to 1.4 s; the enclosed passenger bus in the foreground covers independently moving adult participant 3 from about 3.4 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..da9d93da622e3815fae548067387fb34c72716da
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek white body and tinted windows, positioned in the background, its form softened by the focus - at the start of the clip its front away from the camera. An open-top buggy with a matte slate body and exposed mechanical details, positioned in the lower right foreground, partially cut off by the frame edge - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking up the entry steps into a stationary passenger bus, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek white body and tinted windows, positioned in the background, its form softened by the focus - at the start of the clip its front away from the camera. An open-top buggy with a matte slate body and exposed mechanical details, positioned in the lower right foreground, partially cut off by the frame edge - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the center foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the left of the protagonist, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing on the far left, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.667, the protagonist walks up the entry steps of the stationary bus; the protagonist moves about 2 metres along its route, rises a short distance; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about 2 metres along its route. From 00:01.667 to 00:05.000, the protagonist moves along the aisle inside the bus; the protagonist moves about 4 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 3 moves about 2 metres along its route; the vehicle moves about 4 metres along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:01.667 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 3 metres from the protagonist. The enclosed passenger bus in the foreground covers the protagonist from about 1.2 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 2 from about 3.8 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.9 s to 1.4 s; the enclosed passenger bus in the foreground covers independently moving adult participant 3 from about 3.4 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..451aad4c3c6703faf3f40b3978f248c383293181
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-135-indoor-busBoard/follow/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 4 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek white body and tinted windows, positioned in the background, its form softened by the focus - at the start of the clip its front away from the camera. An open-top buggy with a matte slate body and exposed mechanical details, positioned in the lower right foreground, partially cut off by the frame edge - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the passenger car; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+<Subject 7> is independently moving adult participant 3; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking up the entry steps into a stationary passenger bus, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek white body and tinted windows, positioned in the background, its form softened by the focus - at the start of the clip its front away from the camera. An open-top buggy with a matte slate body and exposed mechanical details, positioned in the lower right foreground, partially cut off by the frame edge - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: An adult office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the center foreground, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is a student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing to the left of the protagonist, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a bicycle courier with cropped black hair, wearing a mustard windbreaker, slate cargo shorts, and ivory sneakers, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 7> (independently moving adult participant 3) is a tourist with curly copper hair, wearing a coral floral shirt, beige shorts, and sandals, standing on the far left, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:01.667, the protagonist walks up the entry steps of the stationary bus; the protagonist moves about 2 metres along its route, rises a short distance; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; the vehicle moves about 2 metres along its route. From 00:01.667 to 00:05.000, the protagonist moves along the aisle inside the bus; the protagonist moves about 4 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 3 moves about 2 metres along its route; the vehicle moves about 4 metres along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:01.667 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 3 metres from the protagonist. The enclosed passenger bus in the foreground covers the protagonist from about 1.2 s to 5.0 s; the enclosed passenger bus in the foreground covers independently moving adult participant 2 from about 3.8 s to 5.0 s; independently moving adult participant 1 in the foreground covers independently moving adult participant 3 from about 0.9 s to 1.4 s; the enclosed passenger bus in the foreground covers independently moving adult participant 3 from about 3.4 s to 5.0 s. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9ae3cbe768d3772cb809fe7b92d48834f01e78f0
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/colored-obb/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 human proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears). Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek, metallic silver body, complete with tinted windows, side mirrors, and heavy rubber tires, positioned in the background - at the start of the clip its front away from the camera. A rugged pickup truck with a matte charcoal body, visible door handles, and a textured bed, positioned in the lower right foreground - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the passenger car: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the object proxy tagged #6636f7 that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the human proxy tagged #36f7d6 that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the human proxy tagged #f79e36 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking along the enclosed aisle of a bus past the seats, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek, metallic silver body, complete with tinted windows, side mirrors, and heavy rubber tires, positioned in the background - at the start of the clip its front away from the camera. A rugged pickup truck with a matte charcoal body, visible door handles, and a textured bed, positioned in the lower right foreground - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the distance on the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks along the enclosed aisle of the bus past the rows of seats; the protagonist moves about 6 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; the vehicle moves about 8 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop in the aisle and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 4 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a3bcf6d38804dd322232c1747c7f7eb0be7b87be
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/mixed/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 blue mannequin proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek, metallic silver body, complete with tinted windows, side mirrors, and heavy rubber tires, positioned in the background - at the start of the clip its front away from the camera. A rugged pickup truck with a matte charcoal body, visible door handles, and a textured bed, positioned in the lower right foreground - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the red box proxy; use the target appearance described below.
+<Subject 4> is the passenger car: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the red box proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the blue mannequin proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the blue mannequin proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking along the enclosed aisle of a bus past the seats, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek, metallic silver body, complete with tinted windows, side mirrors, and heavy rubber tires, positioned in the background - at the start of the clip its front away from the camera. A rugged pickup truck with a matte charcoal body, visible door handles, and a textured bed, positioned in the lower right foreground - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the distance on the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks along the enclosed aisle of the bus past the rows of seats; the protagonist moves about 6 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; the vehicle moves about 8 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop in the aisle and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 4 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/white/prompt.txt b/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ccbf3d4fb00770dddf39e2b196f8c33a313000bc
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-136-indoor-busInside/first-person/white/prompt.txt
@@ -0,0 +1,44 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 2 grey block-figure proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2) and 2 object proxies standing for <Subject 3> (the enclosed passenger bus), <Subject 4> (the passenger car); nothing else in <Video 1> is a person or an interaction object. <Subject 4> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek, metallic silver body, complete with tinted windows, side mirrors, and heavy rubber tires, positioned in the background - at the start of the clip its front away from the camera. A rugged pickup truck with a matte charcoal body, visible door handles, and a textured bed, positioned in the lower right foreground - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the enclosed passenger bus; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the passenger car: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 5> is independently moving adult participant 1; in <Video 1> the grey block-figure proxy that starts on the right of the frame; use the target appearance described below.
+<Subject 6> is independently moving adult participant 2; in <Video 1> the grey block-figure proxy that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is walking along the enclosed aisle of a bus past the seats, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A transit station with platforms, a stationary passenger bus, gates and waiting areas. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A large coach bus with a sleek, metallic silver body, complete with tinted windows, side mirrors, and heavy rubber tires, positioned in the background - at the start of the clip its front away from the camera. A rugged pickup truck with a matte charcoal body, visible door handles, and a textured bed, positioned in the lower right foreground - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: a realistic adult protagonist in practical contemporary clothing rendered in this photographic manner, with a natural face, hair, skin, fabric and footwear; keep this identity and outfit consistent throughout the shot (first-person view from the protagonist's own eyes: the protagonist's body is never visible in this camera; only hand-held items, the mount or the vehicle interior belonging to the protagonist can appear). <Subject 3> is the enclosed passenger bus, realized with the materials above. <Subject 4> is the passenger car, realized with the materials above. <Subject 5> (independently moving adult participant 1) is an office worker with neat short chestnut hair, wearing an ivory shirt, charcoal suit trousers, and polished oxford shoes, standing in the distance on the right, seen only from behind - at the start of the clip facing away from the camera, and keeps exactly this appearance. <Subject 6> (independently moving adult participant 2) is a tourist with curly copper hair under a straw hat, wearing a coral floral shirt, beige shorts, and sandals, standing further back on the left - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.767, the protagonist walks along the enclosed aisle of the bus past the rows of seats; the protagonist moves about 6 metres along its route, now stands on the enclosed passenger bus; independently moving adult participant 1 moves about 3 metres along its route; independently moving adult participant 2 moves about 5 metres along its route; the vehicle moves about 8 metres along its route. From 00:04.767 to 00:05.000, the protagonist comes to a stop in the aisle and stands still; the vehicle moves about half a metre along its route. Recorded relations: the protagonist inside the enclosed passenger bus from 00:00.000 to 00:05.000. The passenger car is one continuous object throughout and ends the clip about 4 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c5f57f72cad2c247e7a3bfd2b0b73809e383518c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/colored-obb/prompt.txt
@@ -0,0 +1,32 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: The protagonist is a realistic adult male engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots. He stands in the center of the frame, facing forward, with a calm and observant demeanor, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fa8b799bea5e405aa7b413c9c1e550520ac8289e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/mixed/prompt.txt
@@ -0,0 +1,32 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: The protagonist is a realistic adult male engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots. He stands in the center of the frame, facing forward, with a calm and observant demeanor, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8395c5649528a22cc791ac2345acb4602c78cf17
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-137-indoor-stairsDown/follow/white/prompt.txt
@@ -0,0 +1,32 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking down stairs and reaching the lower landing. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. <Subject 1>, the protagonist, has the following target appearance: The protagonist is a realistic adult male engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots. He stands in the center of the frame, facing forward, with a calm and observant demeanor, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.667, the protagonist walks down the stairs step by step; the protagonist moves about 5 metres along its route, drops about a metre. From 00:04.667 to 00:05.000, the protagonist reaches the lower landing and walks on; the protagonist is back on the ground. Keep the reference's fixed 53-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..18ffcaa21db5b14cfa4178da8ae3db2141c9db60
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/colored-obb/prompt.txt
@@ -0,0 +1,54 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 1> (the protagonist), <Subject 12> (the training opponent moving laterally and returning fire), <Subject 13> (the second training opponent returning fire from the opposite side) and 5 object proxies standing for <Subject 3> (the solid waist-height cover used in both exchanges), <Subject 4> (the solid cover that stops incoming shots), <Subject 5> (the training rifle with stock, receiver, barrel and magazine), <Subject 8> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 9> is not in frame at 00:00.000 and enters at about 0.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid, waist-high industrial crate made of dark matte composite material, sitting firmly on the floor. A sleek training rifle with a matte black receiver, a long barrel, and a detachable magazine, held in a two-handed grip. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the solid waist-height cover used in both exchanges; in <Video 1> the solid proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid cover that stops incoming shots; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 5> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 6> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 7> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 8> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 9> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the solid proxy shape; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the orange flame shape that appears when it is active; use the target appearance described below.
+<Subject 12> is the training opponent moving laterally and returning fire; in <Video 1> the human proxy tagged #36f7d6; use the target appearance described below.
+<Subject 13> is the second training opponent returning fire from the opposite side; in <Video 1> the human proxy tagged #f79e36; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 13>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 9> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid, waist-high industrial crate made of dark matte composite material, sitting firmly on the floor. A sleek training rifle with a matte black receiver, a long barrel, and a detachable magazine, held in a two-handed grip. <Subject 1>, the protagonist, has the following target appearance: A site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in a ready stance, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 4> is the solid cover that stops incoming shots, realized with the materials above. <Subject 5> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 6> is the short-lived shot tracer, realized with the materials above. <Subject 7> is the short-lived muzzle flash, realized with the materials above. <Subject 8> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 9> is the short-lived shot tracer, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (the training opponent moving laterally and returning fire) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in a combat stance - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (the second training opponent returning fire from the opposite side) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in a combat stance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; the training opponent moving laterally and returning fire moves about half a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), lowers the hands, straightens the legs; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 5> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 8> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. <Subject 5> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 8> is one continuous object throughout and comes to rest about 6 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 0.5 s to 1.0 s; the protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 0.5 s to 1.0 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 0.6 s to 1.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..950ac0a5c4b7294ffa9330c9da9c6d506bb702fd
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/mixed/prompt.txt
@@ -0,0 +1,54 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 blue mannequin proxies standing for <Subject 1> (the protagonist), <Subject 12> (the training opponent moving laterally and returning fire), <Subject 13> (the second training opponent returning fire from the opposite side) and 5 object proxies standing for <Subject 3> (the solid waist-height cover used in both exchanges), <Subject 4> (the solid cover that stops incoming shots), <Subject 5> (the training rifle with stock, receiver, barrel and magazine), <Subject 8> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 9> is not in frame at 00:00.000 and enters at about 0.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid, waist-high industrial crate made of dark matte composite material, sitting firmly on the floor. A sleek training rifle with a matte black receiver, a long barrel, and a detachable magazine, held in a two-handed grip. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the solid waist-height cover used in both exchanges; in <Video 1> the black solid obstacle proxy shape that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid cover that stops incoming shots; in <Video 1> the black solid obstacle proxy shape; use the target appearance described below.
+<Subject 5> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 6> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 7> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 8> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 9> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the black solid weapon proxy shape; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the orange flame shape (#e9aa47) that appears when it is active; use the target appearance described below.
+<Subject 12> is the training opponent moving laterally and returning fire; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 13> is the second training opponent returning fire from the opposite side; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 13>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 9> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid, waist-high industrial crate made of dark matte composite material, sitting firmly on the floor. A sleek training rifle with a matte black receiver, a long barrel, and a detachable magazine, held in a two-handed grip. <Subject 1>, the protagonist, has the following target appearance: A site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in a ready stance, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 4> is the solid cover that stops incoming shots, realized with the materials above. <Subject 5> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 6> is the short-lived shot tracer, realized with the materials above. <Subject 7> is the short-lived muzzle flash, realized with the materials above. <Subject 8> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 9> is the short-lived shot tracer, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (the training opponent moving laterally and returning fire) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in a combat stance - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (the second training opponent returning fire from the opposite side) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in a combat stance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; the training opponent moving laterally and returning fire moves about half a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), lowers the hands, straightens the legs; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 5> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 8> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. <Subject 5> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 8> is one continuous object throughout and comes to rest about 6 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 0.5 s to 1.0 s; the protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 0.5 s to 1.0 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 0.6 s to 1.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3d7dc9fd3a78a856825da9d2385671154b6c92d5
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-139-indoor-combat/follow/white/prompt.txt
@@ -0,0 +1,54 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 1> (the protagonist), <Subject 12> (the training opponent moving laterally and returning fire), <Subject 13> (the second training opponent returning fire from the opposite side) and 5 object proxies standing for <Subject 3> (the solid waist-height cover used in both exchanges), <Subject 4> (the solid cover that stops incoming shots), <Subject 5> (the training rifle with stock, receiver, barrel and magazine), <Subject 8> (the training rifle with stock, receiver, barrel and magazine), <Subject 10> (the training rifle with stock, receiver, barrel and magazine); nothing else in <Video 1> is a person or an interaction object. <Subject 6> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 7> is not in frame at 00:00.000 and enters at about 0.3 s; <Subject 9> is not in frame at 00:00.000 and enters at about 0.4 s; <Subject 11> is not in frame at 00:00.000 and enters at about 0.6 s.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid, waist-high industrial crate made of dark matte composite material, sitting firmly on the floor. A sleek training rifle with a matte black receiver, a long barrel, and a detachable magazine, held in a two-handed grip. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the solid waist-height cover used in both exchanges; in <Video 1> the plain grey proxy solid that starts on the right of the frame; use the target appearance described below.
+<Subject 4> is the solid cover that stops incoming shots; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 6> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 7> is the short-lived muzzle flash: not in frame at 00:00.000; it enters the frame at about 0.3 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 8> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 9> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.4 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 10> is the training rifle with stock, receiver, barrel and magazine; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 11> is the short-lived shot tracer: not in frame at 00:00.000; it enters the frame at about 0.6 s; in <Video 1> the grey flame-shaped solid that appears when it is active; use the target appearance described below.
+<Subject 12> is the training opponent moving laterally and returning fire; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 13> is the second training opponent returning fire from the opposite side; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> aiming and firing at a moving training opponent, then crouching behind cover, involving <Subject 3> to <Subject 13>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 6> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 7> (enters [Shot 1] at about 0.3 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 8> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 9> (enters [Shot 1] at about 0.4 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 10> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 11> (enters [Shot 1] at about 0.6 s): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object with the target appearance described below, one continuous object once it appears.
+<Subject 12> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 13> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A roofed warehouse with tall stocked racks, wide cross aisles, crates and handcarts. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles, animals and objects in this shot: A solid, waist-high industrial crate made of dark matte composite material, sitting firmly on the floor. A sleek training rifle with a matte black receiver, a long barrel, and a detachable magazine, held in a two-handed grip. <Subject 1>, the protagonist, has the following target appearance: A site engineer with short wavy dark hair, wearing a cream polo shirt, slate trousers, and brown boots, standing in a ready stance, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the solid waist-height cover used in both exchanges, realized with the materials above. <Subject 4> is the solid cover that stops incoming shots, realized with the materials above. <Subject 5> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 6> is the short-lived shot tracer, realized with the materials above. <Subject 7> is the short-lived muzzle flash, realized with the materials above. <Subject 8> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 9> is the short-lived shot tracer, realized with the materials above. <Subject 10> is the training rifle with stock, receiver, barrel and magazine, realized with the materials above. <Subject 11> is the short-lived shot tracer, realized with the materials above. <Subject 12> (the training opponent moving laterally and returning fire) is a security guard with a shaved head, wearing a slate uniform shirt with epaulettes, charcoal uniform trousers, and black duty boots, standing in a combat stance - at the start of the clip facing toward the camera, and keeps exactly this appearance. <Subject 13> (the second training opponent returning fire from the opposite side) is a mechanic with auburn hair tied back, wearing navy work overalls over a cream tee and steel-toe boots, standing in a combat stance - at the start of the clip facing toward the camera, and keeps exactly this appearance. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.567, the protagonist raises the training firearm and aims at the moving training opponent; the protagonist moves about half a metre along its route; the training opponent moving laterally and returning fire moves a short distance along its route; the effect moves about half a metre along its route. From 00:00.567 to 00:01.567, the protagonist fires at the training opponents while they move and return fire; the training opponent moving laterally and returning fire moves about half a metre along its route. From 00:01.567 to 00:02.167, the protagonist runs to the nearby cover; the protagonist moves about one and a half metres along its route, lowers the body by about half a metre (head height 1.6 to 1.0 m), bends the knees deeply; the effect moves about one and a half metres along its route, drops about half a metre. From 00:02.167 to 00:02.867, the protagonist crouches behind the cover and reloads the training firearm. From 00:02.867 to 00:03.433, the protagonist rises from behind the cover and steps back out; the protagonist moves about 2 metres along its route, rises to full height (head height 0.9 to 1.6 m), lowers the hands, straightens the legs; the effect moves about 2 metres along its route, rises about half a metre. From 00:03.433 to 00:04.533, the protagonist fires at the training opponents again while they keep moving and returning fire; the protagonist moves a short distance along its route; the training opponent moving laterally and returning fire moves about half a metre along its route; the second training opponent returning fire from the opposite side moves about half a metre along its route; the effect moves a short distance along its route. From 00:04.533 to 00:05.000, the protagonist lowers the training firearm and stands still. Recorded relations: the protagonist holds <Subject 5> from 00:00.000 to 00:05.000; the training opponent moving laterally and returning fire holds <Subject 8> from 00:00.000 to 00:05.000; the second training opponent returning fire from the opposite side holds <Subject 10> from 00:00.000 to 00:05.000. <Subject 5> is one continuous object throughout and ends the clip about half a metre from the protagonist. <Subject 8> is one continuous object throughout and comes to rest about 6 metres from the protagonist. <Subject 10> is one continuous object throughout and ends the clip about 8 metres from the protagonist. The protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 0.5 s to 1.0 s; the protagonist in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 0.5 s to 1.0 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 1.6 s to 1.9 s; the short-lived shot tracer in the foreground covers the second training opponent returning fire from the opposite side from about 0.6 s to 1.0 s. Keep the reference's recorded vertical field of view at every frame. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..782b8bba0f6218ae134456ed2929c77a16c1a1cd
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a golden-brown coat, standing alert on the damp grass, its fur slightly matted from the morning dew, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A small dog with a golden-brown coat, walking on the pasture, its ears perked up and eyes reflecting the soft light, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small dog; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a golden-brown coat, standing alert on the damp grass, its fur slightly matted from the morning dew, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A small dog with a golden-brown coat, walking on the pasture, its ears perked up and eyes reflecting the soft light, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, seen only from behind, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9f0c504054af5ca515eb9d83a2c7aaf58dc12a8e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a golden-brown coat, standing alert on the damp grass, its fur slightly matted from the morning dew, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A small dog with a golden-brown coat, walking on the pasture, its ears perked up and eyes reflecting the soft light, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small dog; in <Video 1> the black solid dog proxy shape; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the black solid dog proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a golden-brown coat, standing alert on the damp grass, its fur slightly matted from the morning dew, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A small dog with a golden-brown coat, walking on the pasture, its ears perked up and eyes reflecting the soft light, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, seen only from behind, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..68941f727343b073dc17bb5b52a816aa7f036f83
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-143-nature-dog-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a golden-brown coat, standing alert on the damp grass, its fur slightly matted from the morning dew, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A small dog with a golden-brown coat, walking on the pasture, its ears perked up and eyes reflecting the soft light, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small dog; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a golden-brown coat, standing alert on the damp grass, its fur slightly matted from the morning dew, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. A small dog with a golden-brown coat, walking on the pasture, its ears perked up and eyes reflecting the soft light, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, seen only from behind, drawn as an anime character in the same sparkly soft romance anime manner (character design: realistically proportioned bodies with simplified anime faces) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — sparkly soft romance anime look: glossy large eyes with light reflections, pastel gradients, floating petals and sparkles, delicate thin lines and dreamy bloom; a monochrome-leaning palette with one accent colour, misty morning backlight, gentle wind moving loose fabric and foliage, painted backgrounds with soft atmospheric depth; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fb757f367845ff6f249ac55a0bd10b51ddf75677
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a light brown coat, standing to the right of the protagonist, its body slightly turned towards the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. Another small dog with a light brown coat, standing to the left of the protagonist, facing slightly away from the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small dog; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a light brown coat, standing to the right of the protagonist, its body slightly turned towards the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. Another small dog with a light brown coat, standing to the left of the protagonist, facing slightly away from the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: The protagonist is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the center of the frame, seen only from behind, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bd98bc270388d4c0b92195a06554e496b5cc201e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a light brown coat, standing to the right of the protagonist, its body slightly turned towards the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. Another small dog with a light brown coat, standing to the left of the protagonist, facing slightly away from the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small dog; in <Video 1> the black solid dog proxy shape; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the black solid dog proxy shape; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a light brown coat, standing to the right of the protagonist, its body slightly turned towards the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. Another small dog with a light brown coat, standing to the left of the protagonist, facing slightly away from the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: The protagonist is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the center of the frame, seen only from behind, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4d9344a42176ca5e33d1ec8a4f4ca6b90c1f7aa0
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-145-nature-dog-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a light brown coat, standing to the right of the protagonist, its body slightly turned towards the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. Another small dog with a light brown coat, standing to the left of the protagonist, facing slightly away from the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small dog; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a light brown coat, standing to the right of the protagonist, its body slightly turned towards the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. Another small dog with a light brown coat, standing to the left of the protagonist, facing slightly away from the camera, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: The protagonist is a hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing in the center of the frame, seen only from behind, painted in the same plein-air oil sketch manner (character design: carefully modelled figures with clear faces) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, plein-air oil sketch: quick loose strokes, sunlit high-key colour, visible underpainting at the edges, canvas texture and painted edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, late-afternoon light through drifting cloud, light drifting haze, a luminous sky taking up much of the frame; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7c272464d7be6a0a486192db696c9eab74c37b8b
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog), <Subject 5> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a smooth, golden-tan coat, standing alert on the paved surface, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, positioned slightly further back on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, standing in the foreground on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small dog; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small dog; in <Video 1> the object proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a smooth, golden-tan coat, standing alert on the paved surface, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, positioned slightly further back on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, standing in the foreground on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing an olive ranger shirt, tan trousers, and brown boots, standing upright in the center of the frame, seen only from behind, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. <Subject 5> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2f1fe29683b0565691a28bccf141cdfd04d3f85a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog), <Subject 5> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a smooth, golden-tan coat, standing alert on the paved surface, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, positioned slightly further back on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, standing in the foreground on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small dog; in <Video 1> the black solid dog proxy shape; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the black solid dog proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small dog; in <Video 1> the black solid dog proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a smooth, golden-tan coat, standing alert on the paved surface, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, positioned slightly further back on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, standing in the foreground on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing an olive ranger shirt, tan trousers, and brown boots, standing upright in the center of the frame, seen only from behind, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. <Subject 5> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..30523c48dc2425a2ca70ca30b926b656b7f9383c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-146-city-dog-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog), <Subject 5> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a smooth, golden-tan coat, standing alert on the paved surface, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, positioned slightly further back on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, standing in the foreground on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small dog; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small dog; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a smooth, golden-tan coat, standing alert on the paved surface, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, positioned slightly further back on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small dog with a smooth, golden-tan coat, standing in the foreground on the left, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing an olive ranger shirt, tan trousers, and brown boots, standing upright in the center of the frame, seen only from behind, rendered as a stylized cartoon 3D character in the same painterly cel-shaded 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. <Subject 5> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly cel-shaded 3D: soft cel shading, watercolour-like distant tones, simplified natural forms and airy light; a rich saturated jewel-tone palette, misty morning backlight, crisp cold air with a hint of frost, glossy smooth surfaces with sharp highlights, busy colourful background props; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..db903e6786f2bd1e1627430ff696eaf68bb54e27
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a tan coat, standing alert with its tail slightly raised, positioned to the right of the protagonist - at the start of the clip its front away from the camera. A small dog with a tan coat, walking with its head turned slightly toward the protagonist, positioned to the left and further back - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small dog; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a tan coat, standing alert with its tail slightly raised, positioned to the right of the protagonist - at the start of the clip its front away from the camera. A small dog with a tan coat, walking with its head turned slightly toward the protagonist, positioned to the left and further back - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing upright with a relaxed posture, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2f286a5c22397592c22883066db858f8aade24e3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a tan coat, standing alert with its tail slightly raised, positioned to the right of the protagonist - at the start of the clip its front away from the camera. A small dog with a tan coat, walking with its head turned slightly toward the protagonist, positioned to the left and further back - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small dog; in <Video 1> the black solid dog proxy shape; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the black solid dog proxy shape; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a tan coat, standing alert with its tail slightly raised, positioned to the right of the protagonist - at the start of the clip its front away from the camera. A small dog with a tan coat, walking with its head turned slightly toward the protagonist, positioned to the left and further back - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing upright with a relaxed posture, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4147c25f2b02abe97b8fe6137af59d70cd9de0d9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-147-city-dog-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (dog) standing for <Subject 3> (the small dog), <Subject 4> (the small dog); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a tan coat, standing alert with its tail slightly raised, positioned to the right of the protagonist - at the start of the clip its front away from the camera. A small dog with a tan coat, walking with its head turned slightly toward the protagonist, positioned to the left and further back - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small dog; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small dog; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while small dogs trot alongside, turn and stop, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small dog with a tan coat, standing alert with its tail slightly raised, positioned to the right of the protagonist - at the start of the clip its front away from the camera. A small dog with a tan coat, walking with its head turned slightly toward the protagonist, positioned to the left and further back - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing upright with a relaxed posture, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small dog, realized with the materials above. <Subject 4> is the small dog, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the dogs trot alongside and turn with it; the protagonist moves about 3 metres along its route; the dog moves about 3 metres along its route; the dog moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the dogs as they stop and look around. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..faa482c4f5f350fdfa4910bfd2a00bcb22495bb4
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a sleek, light beige coat, standing alert on the mossy ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, positioned slightly further back in the mid-ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, walking in the foreground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small cat; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the object proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a sleek, light beige coat, standing alert on the mossy ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, positioned slightly further back in the mid-ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, walking in the foreground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing upright in the center of the frame, seen only from behind, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..712153370506ca1fc1f8ebd782236cf38fbe787a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a sleek, light beige coat, standing alert on the mossy ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, positioned slightly further back in the mid-ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, walking in the foreground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small cat; in <Video 1> the black solid cat proxy shape; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the black solid cat proxy shape; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the black solid cat proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a sleek, light beige coat, standing alert on the mossy ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, positioned slightly further back in the mid-ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, walking in the foreground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing upright in the center of the frame, seen only from behind, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b3d767fa69f25da166305cd859188349cc1575b7
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-154-nature-cat-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a sleek, light beige coat, standing alert on the mossy ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, positioned slightly further back in the mid-ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, walking in the foreground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small cat; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a sleek, light beige coat, standing alert on the mossy ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, positioned slightly further back in the mid-ground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small cat with a sleek, light beige coat, walking in the foreground, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, standing upright in the center of the frame, seen only from behind, drawn in the same retro rubber-hose cartoon manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, retro rubber-hose cartoon look: bendy tube-like limbs, thick black outlines, limited warm palette with film grain and a faint vignette, drawn over the full 3D world so every recorded shape stays readable; a cool blue-grey palette with muted warm accents, harsh midday sun with crisp shadows, crisp cold air with a hint of frost, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9a3c211d16f4feec819f4ca0f438a0de0dca102c
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a beige coat, walking on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. Another small cat with a beige coat, walking slightly further back on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small cat; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a beige coat, walking on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. Another small cat with a beige coat, walking slightly further back on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, stands in the center of the room, seen only from behind, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..90bf7d53d33e34a7f8806a233258a53dc3964f06
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a beige coat, walking on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. Another small cat with a beige coat, walking slightly further back on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small cat; in <Video 1> the black solid cat proxy shape; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the black solid cat proxy shape; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a beige coat, walking on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. Another small cat with a beige coat, walking slightly further back on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, stands in the center of the room, seen only from behind, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a4eb567494ccb0baf51263fc1c189b666ba1babe
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-155-indoor-cat-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a beige coat, walking on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. Another small cat with a beige coat, walking slightly further back on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small cat; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous anime-style shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A fully roofed furnished home with connected rooms, cabinets, chairs and tables. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a beige coat, walking on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. Another small cat with a beige coat, walking slightly further back on the oak plank floor, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, stands in the center of the room, seen only from behind, drawn as an anime character in the same flat simplified anime manner (character design: compact stocky figures with thick outlines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette, and cel shading never reproduces the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Anime-style rendering: a 2D anime look with hand-drawn character design and painted backgrounds over the full 3D world — flat simplified anime look: big simple shapes, minimal shading, uniform thin outlines, cheerful bright flat colours and plain backgrounds; a low-saturation earthy palette, daylight through tall windows with soft shadows, still clean air; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8bd9f498813e668bcabbe4ee2b31b9ea4b52d3a9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small cat; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the object proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5843cd0bb5042a41ea1d7cc78e7d342cac2316e1
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small cat; in <Video 1> the black solid cat proxy shape; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the black solid cat proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the black solid cat proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d1c34ecdbd34f2546df904ad33cb6e7b5474c339
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-156-indoor-cat-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small cat; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An enclosed public hall with columns, exhibit plinths, theatre steps and seating. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. A small cat with a coat of warm ochre, standing on the polished floor, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, seen only from behind, rendered as a stylized cartoon 3D character in the same painterly stylised 3D manner (character design: big heads on small bodies with stubby limbs) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — painterly stylised 3D: hand-painted-looking textures, exaggerated organic shapes, vivid tropical colours and dramatic skies; a low-saturation earthy palette, cool even fluorescent ceiling light, faint dust motes in the light beams, visible painted brush texture on surfaces, backgrounds in slightly lower saturation than the characters; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..034ce8a3af5f90a5dbd749019d25ff4e1d54252b
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a cream-colored coat, standing on the wet pavement to the right of the student - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing further back on the left side of the plaza - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing in the foreground on the left, facing away from the camera - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small cat; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the object proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a cream-colored coat, standing on the wet pavement to the right of the student - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing further back on the left side of the plaza - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing in the foreground on the left, facing away from the camera - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the plaza, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..28cbea1e867c41b435abdbd7b69b712d3e411f58
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a cream-colored coat, standing on the wet pavement to the right of the student - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing further back on the left side of the plaza - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing in the foreground on the left, facing away from the camera - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small cat; in <Video 1> the black solid cat proxy shape; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the black solid cat proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the black solid cat proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a cream-colored coat, standing on the wet pavement to the right of the student - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing further back on the left side of the plaza - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing in the foreground on the left, facing away from the camera - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the plaza, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..88e4e0643de9e60008814b72cd0e31a33ac15cb0
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-158-city-cat-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (cat) standing for <Subject 3> (the small cat), <Subject 4> (the small cat), <Subject 5> (the small cat); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a cream-colored coat, standing on the wet pavement to the right of the student - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing further back on the left side of the plaza - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing in the foreground on the left, facing away from the camera - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small cat; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small cat; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small cat; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while cats weave across the route, look back and circle round, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small cat with a cream-colored coat, standing on the wet pavement to the right of the student - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing further back on the left side of the plaza - at the start of the clip its front away from the camera. A small cat with a cream-colored coat, standing in the foreground on the left, facing away from the camera - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the center of the plaza, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small cat, realized with the materials above. <Subject 4> is the small cat, realized with the materials above. <Subject 5> is the small cat, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the cats weave across it and circle round; the protagonist moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route; the cat moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the cats as they pause and look back. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..df8ef6c1d93d83c486ad1b7ed860cd1a3593fdd8
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit), <Subject 5> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, walking nearby, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small rabbit; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small rabbit; in <Video 1> the object proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, walking nearby, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing upright in the center of the frame, seen only from behind, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. <Subject 5> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1fc16daf3a4e75f27a39a67badaff4c7941864af
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit), <Subject 5> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, walking nearby, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small rabbit; in <Video 1> the black solid rabbit proxy shape; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the black solid rabbit proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small rabbit; in <Video 1> the black solid rabbit proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, walking nearby, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing upright in the center of the frame, seen only from behind, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. <Subject 5> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3b818d650db906fa4c3cfa6366857265f063315b
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-162-nature-rabbit-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit), <Subject 5> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, walking nearby, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small rabbit; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small rabbit; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, in a short hop, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. A small rabbit with soft, pale cream fur, walking nearby, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A field explorer with long braided dark hair, wearing a khaki field shirt with rolled sleeves, olive cargo trousers, and brown leather boots, standing upright in the center of the frame, seen only from behind, rendered as a stylized cartoon 3D character in the same clean heroic stylised 3D manner (character design: muscular exaggerated heroic builds with tiny feet) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. <Subject 5> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — clean heroic stylised 3D: smooth idealised shapes, crisp readable colour blocks, glossy highlights and bright rim light; a sun-bleached faded palette, soft diffuse studio-like light, light drifting haze, soft airbrushed shading, painted storybook scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3014322c877203304084f4daa47d18bd2d63fb9d
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the right of the dancer - at the start of the clip its front away from the camera. A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the left of the dancer, slightly further back - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small rabbit; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the right of the dancer - at the start of the clip its front away from the camera. A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the left of the dancer, slightly further back - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the center of the plaza, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..43ac19f48d835eb98b7c4cec7689976c9e658587
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the right of the dancer - at the start of the clip its front away from the camera. A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the left of the dancer, slightly further back - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small rabbit; in <Video 1> the black solid rabbit proxy shape; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the black solid rabbit proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the right of the dancer - at the start of the clip its front away from the camera. A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the left of the dancer, slightly further back - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the center of the plaza, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..24e6f59d6bdc76e73be03ee9f9c048f909ecff8a
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-163-city-rabbit-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the right of the dancer - at the start of the clip its front away from the camera. A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the left of the dancer, slightly further back - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small rabbit; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the right of the dancer - at the start of the clip its front away from the camera. A small rabbit with soft, creamy-white fur, in a short hop above the wet pavement to the left of the dancer, slightly further back - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the center of the plaza, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..88c7a9a84328180be1d29d055ae0ce117c78635b
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the right of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the left of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small rabbit; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the right of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the left of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing upright in the center of the street, seen only from behind, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..05bc7c6bed9a14a6647f8b43811069c01ef426b4
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the right of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the left of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small rabbit; in <Video 1> the black solid rabbit proxy shape; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the black solid rabbit proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the right of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the left of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing upright in the center of the street, seen only from behind, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..62ef3014bedab84090552e8f90a61f6a4f363182
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-165-city-rabbit-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the right of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the left of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small rabbit; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the right of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. A small rabbit with creamy beige fur and long ears, in a short hop above the asphalt to the left of the protagonist, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young adult student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing upright in the center of the street, seen only from behind, rendered as a stylized cartoon 3D character in the same storybook gouache-textured 3D manner (character design: tall thin elegant figures with pointed features) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — storybook gouache 3D look: hand-painted opaque textures, soft rounded shapes, warm illustrated colours and gentle vignette light; a rich saturated jewel-tone palette, harsh midday sun with crisp shadows, still humid air after rain, glossy smooth surfaces with sharp highlights, simplified rounded scenery; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d79c033a4eba3463b8e596e01ea3bb1e025c2686
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit), <Subject 5> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, walking on the damp grass with its ears perked up - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small rabbit; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small rabbit; in <Video 1> the object proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, walking on the damp grass with its ears perked up - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A middle-aged farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing upright in the center of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. <Subject 5> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..be5b0f39582d162a148e3245ab33d769b4336e6e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit), <Subject 5> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, walking on the damp grass with its ears perked up - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small rabbit; in <Video 1> the black solid rabbit proxy shape; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the black solid rabbit proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small rabbit; in <Video 1> the black solid rabbit proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, walking on the damp grass with its ears perked up - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A middle-aged farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing upright in the center of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. <Subject 5> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3ee51f7f89bdf817ce4086606c36285b18cb2b26
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-166-nature-rabbit-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (rabbit) standing for <Subject 3> (the small rabbit), <Subject 4> (the small rabbit), <Subject 5> (the small rabbit); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, walking on the damp grass with its ears perked up - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small rabbit; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small rabbit; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small rabbit; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while rabbits remain nearby with intermittent short hops, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, in a short hop above the damp grass with its ears perked up - at the start of the clip its front away from the camera. A small rabbit with dense, creamy-white fur, walking on the damp grass with its ears perked up - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A middle-aged farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing upright in the center of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small rabbit, realized with the materials above. <Subject 4> is the small rabbit, realized with the materials above. <Subject 5> is the small rabbit, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the rabbits hop alongside in short bounds; the protagonist moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route; the rabbit moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the rabbits as they remain nearby, with intermittent short hops. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7de1590d7855e96a202b570a7df6bf8bc27f505b
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird), <Subject 5> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small bird; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small bird; in <Video 1> the object proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the center of the plaza, seen only from behind, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. <Subject 5> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..00811d84c0f723fb31c97747e603851ddf746f65
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird), <Subject 5> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small bird; in <Video 1> the black solid bird proxy shape; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the black solid bird proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small bird; in <Video 1> the black solid bird proxy shape; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the center of the plaza, seen only from behind, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. <Subject 5> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5bdb23c990c85995c6d33264030d42ed4e4da4db
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-172-city-bird-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird), <Subject 5> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small bird; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small bird; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A landscaped public park with branching paths, benches, trees and a paved plaza. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small bird with sleek, dark plumage, captured in mid-flight with wings spread, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A street dancer with bleached short hair, wearing an oversized lilac sweatshirt, black track trousers, and chunky ivory trainers, standing in the center of the plaza, seen only from behind, painted in the same East Asian ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. <Subject 5> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, sumi-e ink wash: visible brush and ink strokes, graded black ink washes with sparse strong colour, rice-paper texture, empty breathing space and hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a monochrome-leaning palette with one accent colour, golden-hour side light with long soft shadows, light drifting haze, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e8d20e87df1ce95c6c77009aa76120aac09ba66d
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small bird; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, stands in the center of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a9680300c040278462dc2a94d2cd8a23adfbf4e5
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small bird; in <Video 1> the black solid bird proxy shape; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the black solid bird proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, stands in the center of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..f19ea2126f9264c3921642dfe57b8097444809cb
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-175-nature-bird-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small bird; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Snow banks, conifers, faceted ice formations and a frozen open path. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. A small bird with beige plumage and a slender beak, captured in mid-flight with wings spread - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A hiker with short dark-brown hair, wearing an olive shell jacket, charcoal trousers, and tan hiking boots, stands in the center of the frame, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8b4cff55c60920f850478fb9d6e7129bde91a3fb
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird), <Subject 5> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small bird; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small bird; in <Video 1> the object proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing a ranger shirt, trousers, and boots, standing upright and facing forward, seen only from behind, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. <Subject 5> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2cc4ee2a687645308f257acff8d860e70ece8c25
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird), <Subject 5> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small bird; in <Video 1> the black solid bird proxy shape; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the black solid bird proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small bird; in <Video 1> the black solid bird proxy shape; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing a ranger shirt, trousers, and boots, standing upright and facing forward, seen only from behind, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. <Subject 5> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..857cf6846d17f161981f4ba722790700e239508b
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-176-nature-bird-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (bird) standing for <Subject 3> (the small bird), <Subject 4> (the small bird), <Subject 5> (the small bird); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small bird; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small bird; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small bird; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous graphic non-photorealistic shot in <Subject 2> of <Subject 1> walking along while birds flap past low and fly in a loose group, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Winding temperate forest trails, large trunks, fallen timber and a woodland camp. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. A small bird with natural plumage, captured in mid-flight with wings spread, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A park ranger with short grey hair, wearing a ranger shirt, trousers, and boots, standing upright and facing forward, seen only from behind, drawn in the same graphite pencil sketch manner (character design: figures in quick gesture lines) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small bird, realized with the materials above. <Subject 4> is the small bird, realized with the materials above. <Subject 5> is the small bird, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the birds flap past low overhead in a loose group; the protagonist moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route; the bird moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the birds as they wheel round and skim past again. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and ends the clip about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Ink and contour lines, flat fills and graphic shadow shapes belong to the drawn characters and objects of the target style; they never trace a proxy box, proxy edge or proxy silhouette and never reproduce the proxy's conditioning colours. Rendering style, stated again because it governs every frame: Graphic non-photorealistic rendering, graphite pencil sketch look: drawn pencil contour lines, cross-hatched and smudged graphite shading, visible paper grain, monochrome grey tones with no colour fills and hand-drawn edges instead of standard 3D shading, drawn over the full 3D world so every recorded shape stays readable; a sun-bleached faded palette, misty morning backlight, clear dry air, background barely sketched in; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a9e1f32367cc862c9100c51a71db7eb5257aa5c3
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (duck) standing for <Subject 3> (the small duck), <Subject 4> (the small duck), <Subject 5> (the small duck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small duck; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small duck; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the small duck; in <Video 1> the object proxy tagged #36f7d6 that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while ducks waddle nearby and turn together, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A middle-aged farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small duck, realized with the materials above. <Subject 4> is the small duck, realized with the materials above. <Subject 5> is the small duck, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the ducks waddle alongside in a line; the protagonist moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the ducks as they turn together and waddle on. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4223a8392491733c519f1a546a25f4b528eccc24
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (duck) standing for <Subject 3> (the small duck), <Subject 4> (the small duck), <Subject 5> (the small duck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small duck; in <Video 1> the black solid duck proxy shape; use the target appearance described below.
+<Subject 4> is the small duck; in <Video 1> the black solid duck proxy shape; use the target appearance described below.
+<Subject 5> is the small duck; in <Video 1> the black solid duck proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while ducks waddle nearby and turn together, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A middle-aged farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small duck, realized with the materials above. <Subject 4> is the small duck, realized with the materials above. <Subject 5> is the small duck, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the ducks waddle alongside in a line; the protagonist moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the ducks as they turn together and waddle on. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..932195f63340a59bbf3d9ea20e4829c9678e6dcf
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-182-nature-duck-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (duck) standing for <Subject 3> (the small duck), <Subject 4> (the small duck), <Subject 5> (the small duck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small duck; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small duck; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the small duck; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> walking along while ducks waddle nearby and turn together, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: Open rolling grassland with fenced paddocks, a barn and grazing horses. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. A small duck with a matte, pale yellow finish, walking on the grass - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A middle-aged farmer with grey hair, wearing a faded denim shirt, ochre canvas trousers, and muddy rubber boots, standing with a relaxed posture, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small duck, realized with the materials above. <Subject 4> is the small duck, realized with the materials above. <Subject 5> is the small duck, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the ducks waddle alongside in a line; the protagonist moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the ducks as they turn together and waddle on. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and comes to rest about 2 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9805aa04c397500b21ddd24de96b76bb397d78c5
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (duck) standing for <Subject 3> (the small duck), <Subject 4> (the small duck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small duck; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small duck; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while ducks waddle nearby and turn together, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle of the street, seen only from behind, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small duck, realized with the materials above. <Subject 4> is the small duck, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the ducks waddle alongside in a line; the protagonist moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the ducks as they turn together and waddle on. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d01b420755ce26f3f78d40b44eafaed04125b570
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (duck) standing for <Subject 3> (the small duck), <Subject 4> (the small duck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small duck; in <Video 1> the black solid duck proxy shape; use the target appearance described below.
+<Subject 4> is the small duck; in <Video 1> the black solid duck proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while ducks waddle nearby and turn together, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle of the street, seen only from behind, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small duck, realized with the materials above. <Subject 4> is the small duck, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the ducks waddle alongside in a line; the protagonist moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the ducks as they turn together and waddle on. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9e35a64c63e6e21d10360b8b6e879af6b4133b06
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-189-city-duck-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (duck) standing for <Subject 3> (the small duck), <Subject 4> (the small duck); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small duck; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small duck; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous stylized-3D shot in <Subject 2> of <Subject 1> walking along while ducks waddle nearby and turn together, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: A residential street with enclosed courtyards, low houses, front steps and parked cars. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. A small duck with a matte, pale ochre finish, standing on the asphalt, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A young student with shoulder-length ash-brown hair, wearing a maroon hoodie, indigo denim jeans, and ivory canvas sneakers, standing in the middle of the street, seen only from behind, rendered as a stylized cartoon 3D character in the same feature-animation caricature 3D manner (character design: squat wide figures with oversized hands) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small duck, realized with the materials above. <Subject 4> is the small duck, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist walks along the route while the ducks waddle alongside in a line; the protagonist moves about 3 metres along its route; the duck moves about 3 metres along its route; the duck moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the ducks as they turn together and waddle on. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and comes to rest about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rounded cartoon shapes and smooth stylised surfaces belong to the target style; they are never the grey, flat-colour or wireframe proxy forms themselves. Rendering style, stated again because it governs every frame: Stylized 3D rendering: a complete, continuous 3D world that clearly departs from photorealism — feature-animation caricature 3D: strong silhouette caricature, soft-shaded painterly materials, muted warm palette; a high-key pastel palette, blue-hour dusk with a few warm practical lights, crisp cold air with a hint of frost, visible painted brush texture on surfaces, busy colourful background props; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..52e9f9b1b3b5f9604de9a13c4c2aa575dd942d9d
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish), <Subject 5> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, streamlined body, rendered in a pale, clear shade that catches the ambient light, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, slightly smaller in scale, hovering near the seagrass with a smooth, rounded form, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, the furthest away, appearing as a delicate, faint shape against the darker background, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small fish; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small fish; in <Video 1> the object proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, streamlined body, rendered in a pale, clear shade that catches the ambient light, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, slightly smaller in scale, hovering near the seagrass with a smooth, rounded form, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, the furthest away, appearing as a delicate, faint shape against the darker background, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, floating in a relaxed, horizontal pose with bare feet, seen only from behind, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. <Subject 5> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance; the fish moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..51cdba0a6dc92bd32c659e1327df1eaf0d83929e
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish), <Subject 5> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, streamlined body, rendered in a pale, clear shade that catches the ambient light, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, slightly smaller in scale, hovering near the seagrass with a smooth, rounded form, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, the furthest away, appearing as a delicate, faint shape against the darker background, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small fish; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the black solid fish proxy shape that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small fish; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, streamlined body, rendered in a pale, clear shade that catches the ambient light, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, slightly smaller in scale, hovering near the seagrass with a smooth, rounded form, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, the furthest away, appearing as a delicate, faint shape against the darker background, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, floating in a relaxed, horizontal pose with bare feet, seen only from behind, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. <Subject 5> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance; the fish moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ea259dd7a32d3d3b41146e4c7b0aef93d936eb72
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-192-underwater-fish-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish), <Subject 5> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, streamlined body, rendered in a pale, clear shade that catches the ambient light, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, slightly smaller in scale, hovering near the seagrass with a smooth, rounded form, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, the furthest away, appearing as a delicate, faint shape against the darker background, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small fish; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+<Subject 5> is the small fish; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, streamlined body, rendered in a pale, clear shade that catches the ambient light, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, slightly smaller in scale, hovering near the seagrass with a smooth, rounded form, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. A small fish, the furthest away, appearing as a delicate, faint shape against the darker background, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A swimmer with short wet dark hair, wearing a navy rash guard and ivory swim shorts, floating in a relaxed, horizontal pose with bare feet, seen only from behind, painted in the same dry-brush textured ink wash manner (character design: figures whose clothes are drawn with long flowing strokes) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. <Subject 5> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance; the fish moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, dry-brush ink wash: rough scratchy brush textures, dark ink accents, warm aged-paper tone with faint gold speckle, calligraphic edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a sun-bleached faded palette, dim deep-blue ambient light, drifting streams of bubbles, short dry scratchy strokes, a few ink-drawn trees or rocks framing the scene; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b5719c88f33c174577359afbad475779c5b033d9
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish), <Subject 5> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient light, swimming near the center of the frame, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering slightly to the left, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a delicate, clear appearance, drifting further back in the mid-ground, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small fish; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the object proxy tagged #f736df; use the target appearance described below.
+<Subject 5> is the small fish; in <Video 1> the object proxy tagged #36f7d6; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient light, swimming near the center of the frame, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering slightly to the left, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a delicate, clear appearance, drifting further back in the mid-ground, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, floating neutrally buoyant in the water, seen only from behind, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. <Subject 5> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance; the fish moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4b720bc62f13f6bfa29453967c1cc0aba76b0408
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish), <Subject 5> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient light, swimming near the center of the frame, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering slightly to the left, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a delicate, clear appearance, drifting further back in the mid-ground, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small fish; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 5> is the small fish; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient light, swimming near the center of the frame, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering slightly to the left, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a delicate, clear appearance, drifting further back in the mid-ground, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, floating neutrally buoyant in the water, seen only from behind, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. <Subject 5> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance; the fish moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d70dad079b512a62bcb4049f31da43ae935ec6a5
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-194-underwater-fish-encounter/follow/white/prompt.txt
@@ -0,0 +1,38 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 3 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish), <Subject 5> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient light, swimming near the center of the frame, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering slightly to the left, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a delicate, clear appearance, drifting further back in the mid-ground, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small fish; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 5> is the small fish; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous painterly shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4> and <Subject 5>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient light, swimming near the center of the frame, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering slightly to the left, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. A small fish with a delicate, clear appearance, drifting further back in the mid-ground, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, floating neutrally buoyant in the water, seen only from behind, painted in the same fine-line ink with delicate colour tints manner (character design: figures as a few confident brush gestures) - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. <Subject 5> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance; the fish moves about 3 metres along its route. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 5> from 00:00.000 to 00:05.000; <Subject 4> and <Subject 5> follow <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. <Subject 5> is one continuous object throughout and ends the clip about 5 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Brushstrokes and painted edges belong to the target style; they never trace a proxy box, proxy edge or proxy silhouette. Rendering style, stated again because it governs every frame: Painterly rendering, fine-line tinted ink look: precise thin ink outlines, delicate transparent colour washes in soft mineral tones, calm even composition on pale silk-like ground, hand-drawn edges instead of standard 3D shading, keeping the full 3D world and every recorded shape readable; a rich saturated jewel-tone palette, bright turquoise shallow-water light, drifting streams of bubbles, distant forms fading to pale grey; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/colored-obb/prompt.txt b/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2861d3f2080514716decb979214a64fe606787ed
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/colored-obb/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 human proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the human proxy tagged #366ef7; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient teal light, swimming near the rocks - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering near the seagrass - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small fish; in <Video 1> the object proxy tagged #a7f736; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the object proxy tagged #f736df that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient teal light, swimming near the rocks - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering near the seagrass - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, swimming horizontally through the water, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/mixed/prompt.txt b/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/mixed/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1eee6854b1cbbc07164d1e0630b9fd0c93cedb41
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/mixed/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide only framewise root position, coarse occupied scale, and whole-body root orientation, which should be followed closely. The proxy figure keeps ONE fixed upright pose for the whole clip: that pose is not an action and carries no information about sitting, crouching, riding, reaching or any limb articulation. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 blue mannequin proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the blue mannequin proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient teal light, swimming near the rocks - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering near the seagrass - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the small fish; in <Video 1> the black solid fish proxy shape; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the black solid fish proxy shape that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and its articulated motion from the timeline below.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient teal light, swimming near the rocks - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering near the seagrass - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, swimming horizontally through the water, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/white/prompt.txt b/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c9137974c05b20ba34d30339f2176fd306baac08
--- /dev/null
+++ b/prompts/unverified_first_frame/dv200-195-underwater-fish-encounter/follow/white/prompt.txt
@@ -0,0 +1,36 @@
+AWM_PROXY_CONTROL. One visual reference accompanies this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. Each subject's position and motion come from <Video 1>. Every subject's appearance comes from the text below, rendered consistently in the worldview and art direction it describes. The whole take is rendered in this style, from the first frame to the last: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+
+Human and animal proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style, and every subject's identity, appearance and materials, follow the text; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. The first frame of the target reproduces frame 0 of <Video 1> in the appearance the text describes, and the clip continues directly from it while performing the timestamped actions.
+
+subject_definitions:
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 1 grey block-figure proxy standing for <Subject 1> (the protagonist) and 2 animal proxies (fish) standing for <Subject 3> (the small fish), <Subject 4> (the small fish); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist; in <Video 1> the grey block-figure proxy; use the target appearance described below.
+<Subject 2> is the environment and its fixed spatial arrangement: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient teal light, swimming near the rocks - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering near the seagrass - at the start of the clip its front away from the camera. In <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the small fish; in <Video 1> the plain grey proxy solid; use the target appearance described below.
+<Subject 4> is the small fish; in <Video 1> the plain grey proxy solid that starts on the left of the frame; use the target appearance described below.
+
+summary:
+[reference generation] Generate a 5.00-second continuous shot in <Subject 2> of <Subject 1> swimming along while fish swim nearby with beating tails and turn as a shoal, involving <Subject 3> and <Subject 4>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and realize the scene with the detailed target materials below.
+
+retention_analysis:
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from the target materials below and render natural human articulation over the recorded block-figure motion.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from the target materials below.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in the target materials below, one continuous object throughout.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] <Subject 2>, the setting: An underwater reef with branching coral, submerged rocks and an enclosed shipwreck. Realize every recorded structure, terrain feature, body of water, vehicle and interaction object with plausible materials appropriate to this setting. Vehicles and animals in this shot: A small fish with a sleek, metallic body reflecting the ambient teal light, swimming near the rocks - at the start of the clip its front away from the camera. A small fish with a streamlined shape and shimmering scales, hovering near the seagrass - at the start of the clip its front away from the camera. <Subject 1>, the protagonist, has the following target appearance: A diver with dark hair hidden, wearing a charcoal wetsuit with slate panels and matching fins, swimming horizontally through the water, seen only from behind - at the start of the clip facing away from the camera; keep this identity and outfit consistent throughout the shot. <Subject 3> is the small fish, realized with the materials above. <Subject 4> is the small fish, realized with the materials above. The person keeps the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:04.567, the protagonist swims forward along the route while the fish swim alongside with beating tails; the protagonist moves about 3 metres along its route, raises the hands to chest height; the fish moves about 3 metres along its route; the fish moves about 3 metres along its route, drops a short distance. From 00:04.567 to 00:05.000, the protagonist slows down and watches the fish as the shoal turns together; the protagonist lowers the hands to the sides. Recorded relations: the protagonist observes and accompanies <Subject 3> from 00:00.000 to 00:05.000; the protagonist observes and accompanies <Subject 4> from 00:00.000 to 00:05.000; <Subject 4> follows <Subject 3> from 00:00.000 to 00:05.000. <Subject 3> is one continuous object throughout and comes to rest about 2 metres from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. Keep the reference's fixed 60-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. Rendering style, stated again because it governs every frame: Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+N/A
+
+non_diegetic_music:
+N/A
diff --git a/prompts/verified_first_frame/dv200-008-city-cart/first-person/colored-obb/prompt.txt b/prompts/verified_first_frame/dv200-008-city-cart/first-person/colored-obb/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a5679c251861ce515b53947285a8b445e1c47f0b
--- /dev/null
+++ b/prompts/verified_first_frame/dv200-008-city-cart/first-person/colored-obb/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide only framewise root position, occupied scale (in this representation the box is the entity's exact extent), and whole-body root orientation, which are exact and must be followed precisely. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors, visual effects, anatomy, bones, joints, limb articulation, detailed posture, torso or head pose, hand pose, action phase, or physical contact. Reconstruct each character's appearance from <Picture 1> and the text, and reconstruct articulated motion from the timestamped text. Never reproduce a human proxy as a skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction. On every human, vehicle and animal proxy in <Video 1> the white face is the front of that entity's body: white there is a direction code, never an object, a material or a light, and it must not be painted. The hex colour tags used below only say which proxy stands for which subject: they are identity codes, never colours or materials of the result. Wherever <Video 1> shows a subject only as a flat single-colour shape, that shape is likewise only a placeholder for position, size and motion: draw the complete real person, animal or object the text names in its place.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by <Picture 1> and the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a view from the back of a wooden cart looking down a wet cobblestone street in a misty old town, where a man in an apron walks on the left, a woman pushes a cart in the center, and a man in a hat walks away on the right, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 human proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the wheeled handcart), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> (the legend colour #366ef7 of this person therefore never appears) and none is visible in <Picture 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a narrow cobblestone street glistening with rain and puddles, flanked by grassy verges and large leafy trees. On the left, a wooden market stall with a flat roof stands near a stone wall. On the right, a black iron streetlamp stands near a large tree trunk. In the background, ochre and beige stone buildings with tiled roofs fade into a thick, cool-toned mist. The sky is a pale, overcast grey-blue; in <Video 1> the same environment layout is represented by flat semantic-colour proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials or conditioning colours.
+<Subject 3> is the wheeled handcart: in <Picture 1> a wooden handcart with a slatted box bed and four spoked wheels, positioned in the immediate foreground, viewed from behind; in <Video 1> the object proxy tagged #a7f736, whose white face is its front.
+<Subject 4> is the wheeled handcart: in <Picture 1> a wooden handcart with a slatted box bed and spoked wheels, being pushed by the woman in the center of the street; in <Video 1> the object proxy tagged #6636f7, whose white face is its front.
+<Subject 5> is independently moving adult participant 1: in <Picture 1> a man with dark hair and a beard, wearing a sage green t-shirt, a tan apron, and tan trousers, walking forward on the left side of the street; in <Video 1> the human proxy tagged #f736df that starts on the left of the frame, whose white face is its front.
+<Subject 6> is independently moving adult participant 2: in <Picture 1> a man wearing a straw hat, a salmon pink patterned short-sleeved shirt, and beige shorts, walking away from the camera on the right side of the street; in <Video 1> the human proxy tagged #36f7d6, whose white face is its front.
+<Subject 7> is independently moving adult participant 3: in <Picture 1> a woman with dark hair, wearing a navy long-sleeved top and white trousers, pushing a wooden handcart in the center of the street; in <Video 1> the human proxy tagged #f79e36, whose white face is its front.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is pushing a handcart forward around a gentle bend and releasing the handle, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, limb articulation, detailed posture and physical contact are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in <Picture 1>, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory, facing (the white face of its box is its front) and interaction role from <Video 1>; realize the object as seen in <Picture 1>, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation (the white face of the box is the front of the body), identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and its articulated motion from the timeline below.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a narrow cobblestone street glistening with rain and puddles, flanked by grassy verges and large leafy trees. On the left, a wooden market stall with a flat roof stands near a stone wall. On the right, a black iron streetlamp stands near a large tree trunk. In the background, ochre and beige stone buildings with tiled roofs fade into a thick, cool-toned mist. The sky is a pale, overcast grey-blue, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is the camera wearer and stays outside the frame: the whole shot is seen through this person's eyes, and whatever the timeline says this person does is experienced as camera motion, never shown as a figure. <Subject 3>, the wheeled handcart, is a wooden handcart with a slatted box bed and four spoked wheels, positioned in the immediate foreground, viewed from behind, as seen in <Picture 1>. <Subject 4>, the wheeled handcart, is a wooden handcart with a slatted box bed and spoked wheels, being pushed by the woman in the center of the street, as seen in <Picture 1>. <Subject 5>, independently moving adult participant 1, is a man with dark hair and a beard, wearing a sage green t-shirt, a tan apron, and tan trousers, walking forward on the left side of the street, as seen in <Picture 1>. <Subject 6>, independently moving adult participant 2, is a man wearing a straw hat, a salmon pink patterned short-sleeved shirt, and beige shorts, walking away from the camera on the right side of the street, as seen in <Picture 1>. <Subject 7>, independently moving adult participant 3, is a woman with dark hair, wearing a navy long-sleeved top and white trousers, pushing a wooden handcart in the center of the street, as seen in <Picture 1>. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist steps up to the handcart and takes hold of its handle; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route; <Subject 4> moves about a metre along its route. From 00:00.967 to 00:04.067, the protagonist pushes the handcart forward around the gentle bend; the protagonist moves about 5 metres along its route; <Subject 3> moves about 5 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; <Subject 4> moves about 3 metres along its route. From 00:04.067 to 00:05.000, the protagonist lets go of the handle and the handcart stays where it stopped; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; <Subject 4> moves about a metre along its route. Recorded relations: the protagonist pushes <Subject 3> from 00:00.967 to 00:04.067. <Subject 3> is one continuous object throughout and comes to rest about a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The rhythmic clatter of wooden wheels on wet cobblestones and the soft squelch of footsteps in puddles, accompanied by the distant, muffled murmur of a waking town.
+
+non_diegetic_music:
+N/A
diff --git a/prompts/verified_first_frame/dv200-008-city-cart/first-person/white/prompt.txt b/prompts/verified_first_frame/dv200-008-city-cart/first-person/white/prompt.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ac95dc51fff598e4bd94f49b68ddbb7261689c13
--- /dev/null
+++ b/prompts/verified_first_frame/dv200-008-city-cart/first-person/white/prompt.txt
@@ -0,0 +1,48 @@
+AWM_PROXY_CONTROL. Two visual references accompany this text. <Video 1> is the conditioning proxy video: treat it as the exact spatiotemporal reference for where every subject is, how large it is and how it moves, but never as an appearance template or literal target geometry. <Picture 1> is the global RGB anchor image: it defines identity and appearance, never motion. <Picture 1> is the ABSOLUTE standard for the worldview, art direction and overall style of the entire take, and on worldview, art direction and style it outranks even the description below. It provides NOTHING about any later frame's camera, composition, motion or interactions. Each subject's position and motion come from <Video 1>. A subject's appearance follows <Picture 1> when that subject is visible in it; subjects not visible in <Picture 1> take their appearance from the text, rendered consistently in <Picture 1>'s worldview and art direction.
+
+Human proxies in <Video 1> provide framewise root position, occupied scale, whole-body root orientation and COARSE limb articulation: the proxy figure's limbs move with the recorded joints, so follow the timing and direction of its limb motion closely, and take everything finer - hand pose, facial expression, grip and physical contact - from the timestamped text. They carry no information about character appearance and do not represent identity, face, hairstyle, clothing, materials, colors or visual effects. Reconstruct each character's appearance from <Picture 1> and the text. Never reproduce a human proxy as a block figure, skeleton, stick figure, mannequin, node-connected body, or exposed rig. Left and right refer to each person's own body unless explicitly stated as screen direction.
+
+Vehicle proxies in <Video 1> provide only framewise position, trajectory, body heading, rotation, and accurate overall length, width, height, and occupied scale; these anchors should be followed closely. They carry no information about vehicle appearance. Never reproduce a vehicle proxy as a wireframe vehicle, cage frame, tubular skeleton, exposed chassis, or low-poly proxy car. Vehicle identity, type, design details, materials, colors, and visual effects are defined by <Picture 1> and the text.
+
+Roads, terrain, buildings, relative depth, rough scale, and layout in <Video 1> provide denser but still approximate structural cues. Conditioning colors, proxy geometry, node markers, rigging structures, and other visualization artifacts must never appear in the final video.
+
+When sources conflict, resolve in this order: (1) geometry, camera, position and motion follow <Video 1>; (2) the worldview, art direction and overall style follow <Picture 1> absolutely - above the description below; identity, appearance and materials follow <Picture 1> for subjects visible in it, and follow the text for subjects not visible in it; (3) the timestamped text defines the concrete events, actions and rewrite targets within those constraints.
+
+Clip contract: this clip is the very beginning of the take - nothing precedes it. <Picture 1> is the global appearance anchor for the whole take, and the first frame of the target is locked to this exact image: the clip begins at precisely this visual state - identities, appearance, materials, art direction, lighting, camera framing and layout - and continues directly from it while performing the timestamped actions. Do not restart, redesign or re-interpret any identity.
+
+Camera: this is a first-person take seen through the protagonist's own eyes. The camera path of <Video 1> is the protagonist's head motion; no proxy of the protagonist is drawn in <Video 1> and the protagonist is never shown as a figure. Only the protagonist's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timestamped text says they reach, hold, push or hand over something.
+
+subject_definitions:
+<Picture 1> is the first frame of [Shot 1]: the photoreal opening frame of the target video, showing a view from the back of a wooden cart looking down a wet cobblestone street in a misty old town, where a man in an apron walks on the left, a woman pushes a cart in the center, and a man in a hat walks away on the right, exactly as they must appear at 00:00.000.
+<Video 1> is the conditioning proxy video of the same 5-second shot (its role is fixed by the control block above). The clip contains exactly 3 grey block-figure proxies standing for <Subject 5> (independently moving adult participant 1), <Subject 6> (independently moving adult participant 2), <Subject 7> (independently moving adult participant 3) and 2 object proxies standing for <Subject 3> (the wheeled handcart), <Subject 4> (the wheeled handcart); nothing else in <Video 1> is a person or an interaction object.
+<Subject 1> is the protagonist, the camera wearer: this is a first-person shot seen through this person's own eyes, so the camera path of <Video 1> is this person's head motion; no box or figure of this person is drawn in <Video 1> and none is visible in <Picture 1>. Never draw this person as a figure in front of the camera; only this person's own hands, forearms or a hand-held item may enter the lower edge of the frame when the timeline says they reach, hold, push or hand over something.
+<Subject 2> is the environment and its fixed spatial arrangement: in <Picture 1> a narrow cobblestone street glistening with rain and puddles, flanked by grassy verges and large leafy trees. On the left, a wooden market stall with a flat roof stands near a stone wall. On the right, a black iron streetlamp stands near a large tree trunk. In the background, ochre and beige stone buildings with tiled roofs fade into a thick, cool-toned mist. The sky is a pale, overcast grey-blue; in <Video 1> the same environment layout is represented by untextured grey proxy geometry: approximate structural cues for layout, relative depth and rough scale, never target materials.
+<Subject 3> is the wheeled handcart: in <Picture 1> a wooden handcart with a slatted box bed and four spoked wheels, positioned in the immediate foreground, viewed from behind; in <Video 1> the plain grey proxy solid.
+<Subject 4> is the wheeled handcart: in <Picture 1> a wooden handcart with a slatted box bed and spoked wheels, being pushed by the woman in the center of the street; in <Video 1> the plain grey proxy solid.
+<Subject 5> is independently moving adult participant 1: in <Picture 1> a man with dark hair and a beard, wearing a sage green t-shirt, a tan apron, and tan trousers, walking forward on the left side of the street; in <Video 1> the grey block-figure proxy that starts on the left of the frame.
+<Subject 6> is independently moving adult participant 2: in <Picture 1> a man wearing a straw hat, a salmon pink patterned short-sleeved shirt, and beige shorts, walking away from the camera on the right side of the street; in <Video 1> the grey block-figure proxy.
+<Subject 7> is independently moving adult participant 3: in <Picture 1> a woman with dark hair, wearing a navy long-sleeved top and white trousers, pushing a wooden handcart in the center of the street; in <Video 1> the grey block-figure proxy.
+
+summary:
+[keyframe completion + reference generation] Starting exactly from <Picture 1> as the first frame of [Shot 1], generate a 5.00-second continuous first-person shot in <Subject 2>, seen through the eyes of <Subject 1>, who is pushing a handcart forward around a gentle bend and releasing the handle, involving <Subject 3> and <Subject 4> and <Subject 5> and <Subject 6> and <Subject 7>. Use <Video 1> for layout, root motion, limb motion, object trajectories, action timing and the camera path, and keep the appearance, materials and lighting established in <Picture 1>.
+
+retention_analysis:
+<Picture 1> ([Shot 1] first frame): fully_preserved - the first frame of the target video reproduces this image exactly, including every person's appearance, the objects, the scene layout, materials, lighting and framing; later frames evolve continuously from it.
+<Video 1> (whole-shot spatial and temporal structure): fully_preserved - geometry, camera, placement, root motion, recorded limb motion and object trajectories follow <Video 1> as a coarse spatiotemporal reference; appearance, materials and surface detail are not encoded in the proxy.
+<Subject 1> (camera wearer, never in frame): partially_preserved - retain the camera path of <Video 1> as this person's own head motion and the interaction timing of the timeline; the body stays outside the frame for the whole shot.
+<Subject 2> (appears in [Shot 1]): partially_preserved - retain terrain, object layout and architectural forms from <Video 1>; surfaces, materials and lighting come from <Picture 1>.
+<Subject 3> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in <Picture 1>, one continuous object throughout.
+<Subject 4> (appears in [Shot 1]): partially_preserved - retain position, scale, trajectory and interaction role from <Video 1>; realize the object as seen in <Picture 1>, one continuous object throughout.
+<Subject 5> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 6> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+<Subject 7> (appears in [Shot 1]): partially_preserved - retain the proxy's framewise root position, occupied scale, root orientation, recorded limb motion, identity continuity and interaction timing from <Video 1>; reconstruct its appearance from <Picture 1> and render natural human articulation over the recorded block-figure motion.
+
+detailed_description:
+Photorealistic rendering with real-camera optics: shallow depth of field that separates the sharp near subject from a softly blurred background, diffuse low-contrast light with soft glowing specular highlights and lifted shadows, cool desaturated blue-grey grading with muted warm amber accents, misty atmospheric perspective softening the distance, wet or glossy reflective surfaces where the setting allows and finely textured matte materials elsewhere, natural human anatomy; no proxy colours, wireframes, block figures or game UI.
+[Shot 1] The shot opens exactly on <Picture 1> and evolves continuously from that frame. <Subject 2>, the setting, is a narrow cobblestone street glistening with rain and puddles, flanked by grassy verges and large leafy trees. On the left, a wooden market stall with a flat roof stands near a stone wall. On the right, a black iron streetlamp stands near a large tree trunk. In the background, ochre and beige stone buildings with tiled roofs fade into a thick, cool-toned mist. The sky is a pale, overcast grey-blue, keeping the spacing and sizes of <Video 1>. <Subject 1>, the protagonist, is the camera wearer and stays outside the frame: the whole shot is seen through this person's eyes, and whatever the timeline says this person does is experienced as camera motion, never shown as a figure. <Subject 3>, the wheeled handcart, is a wooden handcart with a slatted box bed and four spoked wheels, positioned in the immediate foreground, viewed from behind, as seen in <Picture 1>. <Subject 4>, the wheeled handcart, is a wooden handcart with a slatted box bed and spoked wheels, being pushed by the woman in the center of the street, as seen in <Picture 1>. <Subject 5>, independently moving adult participant 1, is a man with dark hair and a beard, wearing a sage green t-shirt, a tan apron, and tan trousers, walking forward on the left side of the street, as seen in <Picture 1>. <Subject 6>, independently moving adult participant 2, is a man wearing a straw hat, a salmon pink patterned short-sleeved shirt, and beige shorts, walking away from the camera on the right side of the street, as seen in <Picture 1>. <Subject 7>, independently moving adult participant 3, is a woman with dark hair, wearing a navy long-sleeved top and white trousers, pushing a wooden handcart in the center of the street, as seen in <Picture 1>. All people keep the same face, build and clothing throughout, with clothing folds responding naturally to movement. The action follows <Video 1> according to this timeline: From 00:00.000 to 00:00.967, the protagonist steps up to the handcart and takes hold of its handle; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about half a metre along its route; independently moving adult participant 3 moves about a metre along its route; <Subject 4> moves about a metre along its route. From 00:00.967 to 00:04.067, the protagonist pushes the handcart forward around the gentle bend; the protagonist moves about 5 metres along its route; <Subject 3> moves about 5 metres along its route; independently moving adult participant 1 moves about 2 metres along its route; independently moving adult participant 2 moves about 3 metres along its route; independently moving adult participant 3 moves about 3 metres along its route; <Subject 4> moves about 3 metres along its route. From 00:04.067 to 00:05.000, the protagonist lets go of the handle and the handcart stays where it stopped; independently moving adult participant 1 moves about half a metre along its route; independently moving adult participant 2 moves about a metre along its route; independently moving adult participant 3 moves about a metre along its route; <Subject 4> moves about a metre along its route. Recorded relations: the protagonist pushes <Subject 3> from 00:00.967 to 00:04.067. <Subject 3> is one continuous object throughout and comes to rest about a metre from the protagonist. <Subject 4> is one continuous object throughout and ends the clip about 3 metres from the protagonist. Keep the reference's fixed 75-degree vertical field of view. Preserve the single take and its pacing, without cuts or a new camera path. Keep world geometry, object identities and materials stable throughout the camera movement; reveal newly visible surfaces with the same layout and appearance; introduce no extra people or foreground obstacles. No interface overlays, captions, watermarks, dialogue or music.
+
+overall_soundscape:
+The rhythmic clatter of wooden wheels on wet cobblestones and the soft squelch of footsteps in puddles, accompanied by the distant, muffled murmur of a waking town.
+
+non_diegetic_music:
+N/A