Upload sample/minih33_smoothloopv3_002000_00_20260919205804_42.mp4.json with huggingface_hub
Browse files
sample/minih33_smoothloopv3_002000_00_20260919205804_42.mp4.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"training_step": 2000,
|
| 3 |
+
"epoch": null,
|
| 4 |
+
"prompt": "For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced. integrated_multimodal_description: [Shot 1] A smooth looping animation in 2D pixel-art style shows one girl seated in the passenger seat of a moving car in a medium interior shot, preserving the character design, seated pose, clothing, car interior layout, and window framing established by <Picture 1>. The girl herself remains almost completely still in the same forward-facing seated posture with her hands resting in her lap throughout, maintaining the opening position while the sense of travel is carried by the environment outside. The cityscape visible through the side and front windows flows continuously past from front to back, with suspension-bridge towers, high-rise buildings, water, and alternating brown tunnel-like wall segments sliding steadily across the glass and being replaced by new sections in a repeating drive cycle. Light color on the windows shifts subtly between bright daylight blue and warmer shadowed tones as different background elements pass. A small hanging ornament near the windshield sways with tiny secondary motion from the car's movement. The car interior, dashboard display, and camera framing stay fixed, making the scrolling exterior clearly distinct from camera motion. The background pattern returns toward a bridge-and-city arrangement similar to the opening view to complete the travel loop. overall_soundscape: N/A non_diegetic_music: N/A",
|
| 5 |
+
"seed": 42,
|
| 6 |
+
"height": 480,
|
| 7 |
+
"width": 832,
|
| 8 |
+
"frames": 124,
|
| 9 |
+
"sigma_points": 20,
|
| 10 |
+
"model_evaluations": 19,
|
| 11 |
+
"keyframe_anchors": [
|
| 12 |
+
"first"
|
| 13 |
+
],
|
| 14 |
+
"metrics": {
|
| 15 |
+
"joint_denoising": {
|
| 16 |
+
"seconds": 101.8772089969134,
|
| 17 |
+
"peak_allocated_gib": 13.180626392364502,
|
| 18 |
+
"peak_reserved_gib": 14.544921875
|
| 19 |
+
},
|
| 20 |
+
"sequential_av_decode": {
|
| 21 |
+
"seconds": 9.143913288018666,
|
| 22 |
+
"peak_allocated_gib": 7.0488786697387695,
|
| 23 |
+
"peak_reserved_gib": 7.94140625
|
| 24 |
+
}
|
| 25 |
+
}
|
| 26 |
+
}
|