stablediffusiontutorials commited on
Commit
3d85cd6
·
verified ·
1 Parent(s): 3e3efed

Upload 4 files

Browse files
MINIMAX_H3_IMAGE_PROMPT_GUIDE.md ADDED
@@ -0,0 +1,419 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # MiniMax H3 Image: Prompt Writing Guide
2
+
3
+ **For people:** drop this file into any LLM (ChatGPT, Claude, Gemini, a local model) and ask it for an image prompt, for example:
4
+
5
+ > *"Using this guide, write me a prompt for a rainy Tokyo street at night with a woman holding a red umbrella."*
6
+ > *"Using this guide, write an edit prompt: Picture 1 is my living room, make it sunset and add a cat on the sofa."*
7
+ > *"Using this guide, put the man from Picture 1 into the café from Picture 2."*
8
+
9
+ The LLM reads the rules below and returns a finished prompt plus the settings to use.
10
+
11
+ **For the LLM:** this whole file is your instruction set. Read it all, then follow section 0.
12
+
13
+ ---
14
+
15
+ ## 0. Instructions for the LLM
16
+
17
+ You write prompts for **MiniMax H3 Image**, a text-to-image and image-editing model. Follow these rules every time.
18
+
19
+ 1. **Pick the mode** (section 2): text-to-image, single-picture edit, or multi-reference composition. The mode decides the template.
20
+ 2. **Lock the aspect ratio first** (section 3.2), from the user's request and the main subject, before you add detail.
21
+ 3. **Decide every detail yourself.** Anything you leave out, the model invents, usually as something generic. Do not ask about small things (hair colour, lens, time of day); choose something that fits and write it down.
22
+ 4. **Ask only when something essential is missing**: what the image should show at all, or, for an edit, what should change. One short question, then stop.
23
+ 5. **Describe what should exist, in visible terms.** There is no negative prompt (section 9). Replace moods and adjectives with things a viewer can point at.
24
+ 6. **Write the prompt in English**, as plain flowing sentences. Keep any text that must appear in the image in its original language, inside double quotes.
25
+ 7. **Run the final check** (section 11) before answering.
26
+ 8. **Answer in this format**, and nothing else besides one or two short comments:
27
+
28
+ ````
29
+ **Mode:** text-to-image | edit | multi-reference
30
+ **Pictures:** Picture 1 = ..., Picture 2 = ... (omit for text-to-image)
31
+
32
+ ```text
33
+ <the finished prompt>
34
+ ```
35
+
36
+ **Settings:** aspect ratio X:Y · draft 4 MP to test, then 16 MP final · Enhance prompt: off
37
+ ````
38
+
39
+ ---
40
+
41
+ ## 1. Model facts
42
+
43
+ | Item | Value |
44
+ |---|---|
45
+ | Model | MiniMax H3 Image (MiniMaxAI/MiniMax-H3-Image), a diffusion transformer |
46
+ | Text encoder | Qwen3-VL 32B, a vision-language LLM. It reads full sentences, spatial words, counts and quoted text well, and it also sees the reference pictures |
47
+ | Sampling | Distilled for **6 steps, CFG 1.0**. More steps do not add detail |
48
+ | Negative prompt | **None.** At CFG 1 the negative branch is zeroed out. Everything is controlled by the positive prompt |
49
+ | Native size | Trained up to **16 MP**. 4 MP is the fast draft tier. Below about 2 MP faces and hands break |
50
+ | Reference pictures | Up to **10**, named `<Picture 1>` ... `<Picture 10>` in the prompt. Picture 1 is the one an edit changes |
51
+ | Prompt length | Long prompts work and help. Typical: 120 to 350 words. Complex scenes with several references: 400 to 900 words |
52
+ | Text rendering | Good for short, quoted text: titles, signs, labels, UI text |
53
+
54
+ ### Sizes (both sides are multiples of 32)
55
+
56
+ | Ratio | Draft 4 MP | Final 16 MP | Use for |
57
+ |---|---|---|---|
58
+ | 1:1 | 2048x2048 | 4096x4096 | icons, products, centred subjects, social posts |
59
+ | 3:4 | 1760x2368 | 3552x4736 | portraits, one standing person, most people shots |
60
+ | 4:3 | 2368x1760 | 4736x3552 | two people side by side, rooms, tabletop |
61
+ | 2:3 | 1664x2496 | 3360x5024 | full-body figure, posters, book covers |
62
+ | 3:2 | 2496x1664 | 5024x3360 | classic photo, street, landscape with a subject |
63
+ | 9:16 | 1536x2720 | 3072x5472 | phone wallpaper, stories, tall buildings |
64
+ | 16:9 | 2720x1536 | 5472x3072 | cinematic scenes, wide action, slides, banners |
65
+ | 21:9 | 3136x1344 | 6272x2688 | panoramas, timelines, genuinely wide banners |
66
+
67
+ **Why 16 MP matters:** at 4 MP, faces far from the camera break (teeth, eyes). At 16 MP, small faces hold together and skin gets real pores. Test the prompt at 4 MP, render the keeper at 16 MP with the same seed.
68
+
69
+ ---
70
+
71
+ ## 2. Choose the mode
72
+
73
+ | The user wants | Mode | Template |
74
+ |---|---|---|
75
+ | A new image from words only | **Text-to-image** | section 4 |
76
+ | To change one existing picture (colour, object, light, background, outfit, text) | **Edit** | section 7 |
77
+ | People, objects, materials or places from several pictures combined into one new image | **Multi-reference** | section 8 |
78
+ | To change a picture *and* put things from other pictures into it | **Edit + references** | section 8.5 |
79
+
80
+ If the user gives pictures but says nothing about changing one, it is multi-reference. If they say "change", "make it", "replace", "remove", "add to this photo", it is an edit.
81
+
82
+ ---
83
+
84
+ ## 3. Before writing: plan silently
85
+
86
+ ### 3.1 Image type
87
+
88
+ Decide which kind of image this is, because it changes what matters most:
89
+
90
+ - **A depicted scene or photograph:** subject identity, pose, contact, scale, environment, light, lens, physical plausibility.
91
+ - **A designed artifact** (poster, cover, infographic, UI, chart, character sheet, product board, packaging, menu): layout, exact text, hierarchy, labels. Plan it as a page, not a scene (section 6).
92
+
93
+ ### 3.2 Aspect ratio
94
+
95
+ Decide in this order. Earlier reasons win.
96
+
97
+ 1. **The user named a ratio or pixel size** → use it (pixels: width ÷ height, closest ratio in the table).
98
+ 2. **The user named a deliverable** → poster/cover/phone wallpaper: vertical. Slide/banner/header/desktop wallpaper: horizontal. Icon/avatar/post: 1:1.
99
+ 3. **The main subject decides**, not the background. One standing person, a tower, stacked information: vertical. Side-by-side people, lateral action, a broad landscape the user asked for: horizontal. Centred, round or compact subjects: 1:1.
100
+ 4. **Weak evidence** → 1:1, 3:4 or 4:3. Use 16:9 and 9:16 only for clearly directional or screen-shaped images, 21:9 only for real panoramas.
101
+
102
+ Words like "cinematic", "portrait", "photorealistic", "dramatic", "wide-angle" do not decide the ratio by themselves.
103
+
104
+ For an **edit**, use the ratio of Picture 1 unless the user asks for another one.
105
+
106
+ ### 3.3 Inventory
107
+
108
+ Write down, for yourself:
109
+
110
+ - **Subjects** with an exact count ("two women", "one dog"), and who is the main one.
111
+ - **Layout:** where each subject sits in the frame (screen-left, centre, screen-right, foreground, middle ground, background) and how big it looks.
112
+ - **Light:** one main source, its direction, colour and hardness.
113
+ - **Camera:** height, angle, lens, distance, focus.
114
+
115
+ Then write the prompt from this plan. Do not change the ratio because of detail you add later.
116
+
117
+ ---
118
+
119
+ ## 4. Text-to-image template
120
+
121
+ Write in this order. Each part is one or more plain sentences.
122
+
123
+ 1. **Medium and shot.** "A candid photograph...", "A close-up product photo...", "A 2D anime illustration...", "A minimalist travel poster...".
124
+ 2. **Main subject.** Count, age (always "adult" for people in any romantic or revealing context), build, face, skin, hair, expression, clothing with colours and materials, accessories.
125
+ 3. **Pose and action.** One frozen moment. Which way the body faces relative to the camera, where the hands are, what they touch (section 5).
126
+ 4. **Other subjects.** Same detail, scaled down for less important ones. Place each one in the frame.
127
+ 5. **Setting.** What is behind, left and right, foreground and background. Concrete objects, materials, wear, weather.
128
+ 6. **Light.** Source, direction, time of day, colour temperature, shadows. "Low warm sun from screen-left, long shadows falling to the right."
129
+ 7. **Camera.** Lens, height, angle, focus. "Shot at eye level with a 50mm lens, f/2.8, background softly out of focus."
130
+ 8. **Texture and finish.** For photos: skin pores, fabric weave, dust, scratches, natural colour. For illustrations: one short style sentence.
131
+ 9. **Text in the image** (if any): exact words in double quotes, where they sit, size, font style, colour.
132
+
133
+ ### 4.1 Photorealism recipe
134
+
135
+ Short prompts give a glossy, retouched "doll face". To get a real photo look:
136
+
137
+ - Write long and physical: age, skin tone, freckles, stubble, flyaway hairs, creases in clothing, a lived-in setting.
138
+ - Name a real capture device and its look. Example: *"Shot on the iPhone 16 Pro front camera, visible skin texture and pores, natural phone HDR, no retouching, ordinary Instagram photo, not cinematic, not studio."* Other good anchors: "35mm film photo, Kodak Portra 400, slight grain", "full-frame mirrorless, 85mm f/1.8", "amateur flash photo at night".
139
+ - Give the light a real source that exists in the scene (a window, a streetlight, the sun) and match its direction on the face and in the cast shadows. Flat, even frontal light makes people look pasted in.
140
+ - State natural colour: "neutral white balance, true-to-life skin tones, no colour grading".
141
+ - Fill real places with life: other people (smaller, mostly seen from behind), parked cars, litter, signs. "Clean" means a sharp image, not an empty street.
142
+ - Skip keyword spam ("8k, masterpiece, ultra HD, trending"). Describe instead.
143
+
144
+ ### 4.2 Other styles
145
+
146
+ Keep the style statement short and specific, then spend the words on content.
147
+
148
+ - Anime: "2D anime illustration, clean line art, soft cel shading."
149
+ - Painting: "Oil painting on canvas, visible thick brush strokes, warm umber underpainting."
150
+ - 3D: "Stylised 3D render, soft subsurface skin, studio three-point lighting, clay-like materials."
151
+ - Fantasy made real: "photographed like a real documentary photo of an impossible event, real textures, not illustration, not CGI."
152
+
153
+ Piling up many style words (gouache + ink + cel + watercolour + ...) muddies the result.
154
+
155
+ ---
156
+
157
+ ## 5. Bodies, direction, hands and scale
158
+
159
+ The model follows spatial language closely, so be exact.
160
+
161
+ ### 5.1 Use camera-relative words
162
+
163
+ - **Positions:** screen-left, screen-right, centre, upper/lower part of the frame, foreground, middle ground, background, "about 12 metres from the camera".
164
+ - **Facing:** front-facing, back-facing (we see the back of the head), left-facing profile, right-facing profile, front three-quarter, rear three-quarter.
165
+ - **Motion:** leftward, rightward, toward the camera, away from the camera into the scene.
166
+
167
+ Never write only "facing the door" or "looking at her friend". Say where the door is in the frame, then which way the body faces: "The door is at screen-right; he walks rightward toward it in a right-facing profile."
168
+
169
+ Check that what is visible matches: walking away shows the back and the back of the head, not the face. Walking toward the camera shows the face and chest.
170
+
171
+ ### 5.2 Spatial order for layered scenes
172
+
173
+ For cover, hiding, behind/in front, state the depth order in one sentence:
174
+ *"Order from the camera: camera, then the man, then the rock wall, then the empty parking lot, then the enemies."*
175
+ Then say what that means: "the wall stands between him and the enemies; we see his back."
176
+
177
+ ### 5.3 Hands and contact
178
+
179
+ - Choose framing that makes the contact readable: waist-up or closer when hands or faces matter. Do not force a full-body view that makes the important detail tiny.
180
+ - Say which hand does what: "her right hand holds the cup handle, left hand flat on the table".
181
+ - Keep hands empty unless something is named.
182
+ - One frozen phase of an action: a limb is either planted or in the air, never both.
183
+
184
+ ### 5.4 Scale and count
185
+
186
+ - Give real sizes when size matters: "about 15 centimetres tall, fitting inside his gloved fist, clearly smaller than his head". Objects taken from a reference picture often come out too large unless you do this.
187
+ - Give figure size for distant people: "each figure about one eighth of the image height".
188
+ - State exact counts: "three people", "five camels". Add "no other people in the frame" when extra people would ruin it.
189
+ - Background crowds: "small figures, mostly seen from behind", so they do not compete with the main subject.
190
+
191
+ ---
192
+
193
+ ## 6. Text, posters, infographics, UI
194
+
195
+ ### 6.1 Text rules
196
+
197
+ - Put every word that must appear in **double quotes**, exactly as it should be spelled, including capitals and punctuation.
198
+ - Say where it sits, how big, font style and colour: *The headline "MIDNIGHT HARVEST" runs across the top in large cream condensed serif capitals.*
199
+ - Each text item appears once. Keep each line short. Long paragraphs on the image become scribbles; use one-line captions.
200
+ - Non-English text stays in its own script inside the quotes: *The sign reads "居酒屋 とり吉".*
201
+
202
+ ### 6.2 Plan designed artifacts as a page
203
+
204
+ Divide the canvas in reading order and fill every slot with real content: header/title, subtitle, hero image, side panels, card rows, chart, diagram, footer. For each slot give its **position**, its **visible text or label**, and **what it shows**.
205
+
206
+ Do not write vague fillers like "icons", "details", "annotations", "supporting graphics". Say what each icon or label shows.
207
+
208
+ Add the small parts that make the artifact feel finished when they fit: tagline, legend, units, dimension labels, rating, quote, QR or logo placeholder, footer note.
209
+
210
+ Specific types:
211
+
212
+ - **Infographic / diagram:** section headers, a main diagram with labels attached to the right parts, a few large numbered steps, thick non-crossing arrows from a named start to a named end, a legend, units.
213
+ - **Chart / dashboard:** title, axis labels, tick values, units, legend entries, data values consistent with the bars or lines.
214
+ - **Character sheet:** name, role, full-body view with readable proportions, front/side/back views if asked, outfit and prop callouts, colour swatches.
215
+ - **Product board:** title, hero render, detail insets, exploded layers, materials, dimension labels, colour swatches.
216
+ - **UI / web page:** navigation, real section names, buttons with labels, fields, cards with plausible values, one selected state.
217
+ - **Poster / cover:** headline, subtitle, one hero subject per category, clear hierarchy, intentional empty space.
218
+
219
+ ---
220
+
221
+ ## 7. Image editing (one picture)
222
+
223
+ Load the picture to change as **Picture 1**. The prompt is an instruction about it.
224
+
225
+ ### 7.1 Edit template
226
+
227
+ ```text
228
+ Edit <Picture 1>: keep <list the things that must not change, by name>.
229
+ <The change, described as what the result looks like, with position, size, colour, material, light.>
230
+ <How the change fits the existing light, perspective and shadows.>
231
+ Everything else in <Picture 1> stays exactly the same.
232
+ ```
233
+
234
+ ### 7.2 Rules
235
+
236
+ 1. **Start with "Edit <Picture 1>:"** so the model knows this is a change, not a new image.
237
+ 2. **Name what to keep, concretely.** Not "keep everything", but: "keep the exact location, camera angle, perspective and daylight of <Picture 1>: the silver truck on the left edge, the cracked asphalt with faded white lines, the brick building in the centre". Listing the visible things locks them in.
238
+ 3. **Keep the person's identity** when a person is in the picture: "keep her exact face, skin tone, hairstyle and expression".
239
+ 4. **Describe the result, not the operation.** "The sofa is now deep green velvet with brass legs" works better than "change sofa colour".
240
+ 5. **Place the change:** where in the frame, how big relative to things already there.
241
+ 6. **Match light and perspective:** "lit by the same warm window light from screen-left, with a soft contact shadow on the floor", "following the same floor perspective".
242
+ 7. **One edit at a time works best.** For three or more unrelated changes, do them as separate edits, or list them clearly as numbered sentences.
243
+ 8. **Turn Enhance prompt off** for edits. The upsampler is for text-to-image only.
244
+ 9. **Keep the ratio of Picture 1.**
245
+
246
+ ### 7.3 Edit recipes
247
+
248
+ | Edit | Example wording |
249
+ |---|---|
250
+ | **Add an object** | *Add a sleeping orange tabby cat curled on the left sofa cushion, about the size of the cushion, lit by the same lamp from screen-right, its shadow falling on the cushion.* |
251
+ | **Remove an object** | *Remove the parked white van on the right. The brick wall and pavement behind it continue naturally where the van was, with the same texture and shadows.* (Always say what fills the space.) |
252
+ | **Replace an object** | *Replace the coffee mug in her right hand with a tall glass of iced matcha latte with visible ice cubes and a paper straw, held the same way.* |
253
+ | **Colour / material** | *The car's paint is now matte olive green; chrome trim, windows, wheels and reflections stay as they are.* |
254
+ | **Outfit** | *Keep her exact face, hair, pose and the room. She now wears a cream cable-knit sweater and dark blue straight jeans.* |
255
+ | **Background** | *Keep the man, his pose, clothing and the camera angle exactly. Replace the background with a misty pine forest at dawn, light coming from behind him to match the rim light on his shoulders.* |
256
+ | **Time of day / weather** | *Same street, same camera. It is now night after rain: wet asphalt reflecting orange streetlights and red tail lights, shop windows glowing warm, a dark blue sky.* |
257
+ | **Relight** | *Keep everything. The only light is now a low golden-hour sun from screen-left, warm on the left side of her face, the right side in soft shadow, long shadows to the right.* |
258
+ | **Expression / pose** | *Keep her identity, outfit and setting. She now laughs with her eyes half closed, head tilted slightly to screen-right, right hand raised to her mouth.* |
259
+ | **Change text** | *The shop sign above the door now reads "BLUE HOUR CAFÉ" in the same white neon script, same size and position.* |
260
+ | **Style conversion** | *Redraw <Picture 1> as a 2D anime illustration with clean line art and soft cel shading. Keep the same composition, poses, outfits and colours.* |
261
+ | **Photo to realistic** | *Turn this 3D render into a real photograph: real human skin with pores, real fabric, natural daylight, true-to-life colour. Keep the face, outfit, pose and framing.* |
262
+ | **Restore / colourise** | *Restore this old photo: remove scratches and dust, sharpen the faces, add natural colour: warm skin tones, a navy coat, a red brick wall. Keep every face and the composition.* |
263
+
264
+ ---
265
+
266
+ ## 8. Multi-reference composition
267
+
268
+ Load pictures in slot order, starting at slot 1. The pictures that are switched on are numbered in that order: with slots 1, 2 and 3 on, they are `<Picture 1>`, `<Picture 2>`, `<Picture 3>`. Always write the angle-bracket form exactly.
269
+
270
+ ### 8.1 Template
271
+
272
+ ```text
273
+ <Subject 1> is the <adult man / woman / object> in <Picture 1>, with exactly the same face, <key visible traits in words: hair, eyes, skin, build>, wearing <the outfit in words>.
274
+ <Subject 2> is the <...> in <Picture 2>, <...>.
275
+ <Picture 3> is the location: <describe it in words>.
276
+ <Picture N> is only a reference for <its face / its outfit / its material / its colours>. Do not carry over its <background / pose / layout / render look>.
277
+
278
+ <Medium and shot, as in section 4: "A candid photograph...">
279
+ <Scene: where each subject is in the frame, facing, action, contact, scale.>
280
+ <Light and colour matching the location picture.>
281
+ <Camera and finish.>
282
+ ```
283
+
284
+ ### 8.2 Rules
285
+
286
+ 1. **Define each subject once, then use the label.** "<Subject 1> is the adult woman in <Picture 1>..." then write "<Subject 1>" in the scene. This prevents faces from swapping or merging.
287
+ 2. **Also describe the reference in words.** "with the same face" plus hair colour, eye colour, outfit and props. Words back up the picture and keep identity from drifting.
288
+ 3. **Say what to take and what to leave.** Character sheets and turnarounds carry grey backgrounds, A-poses, panel layouts and a 3D game look. Write: *"<Picture 1> and <Picture 2> are character sheets. Use only their faces, hair and outfits. Do not carry over their grey studio backgrounds, panel layouts, front-facing A-poses or 3D render look."*
289
+ 4. **Location pictures:** describe what is in them (left side, centre, right side, background), and say the light and colour match it: *"The colours and light match <Picture 3> exactly: neutral daylight, warm ochre walls, pale blue sky, no teal or green colour cast."*
290
+ 5. **Material or style pictures:** *"a waist-high wall made from the material of <Picture 3>: dark fractured rock with glowing orange cracks"*.
291
+ 6. **Object pictures:** give the real-world size and how it is held (section 5.4). Objects lifted from product shots tend to come out oversized.
292
+ 7. **Real people in a scene:** "Nobody poses and nobody looks at the camera" gives a candid result. Mention natural skin texture, sweat, dust, fabric wear.
293
+ 8. **Each picture costs time and memory.** Use only the pictures that add something.
294
+ 9. **Enhance prompt off.**
295
+
296
+ ### 8.3 Common jobs
297
+
298
+ | Job | Key wording |
299
+ |---|---|
300
+ | Person into a new place (no location picture) | *<Subject 1> is the woman in <Picture 1>, same face and hair. She sits at a window table in a small Lisbon café...* |
301
+ | Person into a pictured place | *...<Picture 2> is the location. Place <Subject 1> at the table on the right, lit by the window light of <Picture 2>.* |
302
+ | Two people together | Define both subjects, then their positions: *<Subject 1> on screen-left in a right-facing three-quarter view, <Subject 2> on screen-right facing her...* |
303
+ | Outfit from one picture onto a person from another | *<Subject 1> is the man in <Picture 1>, same face and build. He wears the exact jacket from <Picture 2>: ... Use only the jacket from <Picture 2>.* |
304
+ | Product into a lifestyle scene | *The bottle is the exact perfume bottle from <Picture 1>, same shape, cap and label text "...", about 12 cm tall, standing on a marble bathroom shelf...* |
305
+ | Consistent character across images | Reuse the same subject definition paragraph, word for word, in every prompt. |
306
+
307
+ ### 8.4 Pose or composition reference
308
+
309
+ To copy a pose from a picture without copying the person: *"<Subject 1> takes exactly the pose of the person in <Picture 2>: ... Use only the pose from <Picture 2>, not their face, clothing or background."* Also describe the pose in words.
310
+
311
+ ### 8.5 Edit + references
312
+
313
+ When one picture is the base to change and others supply things to insert:
314
+
315
+ - Base picture is **Picture 1**. Start with *"Edit <Picture 1>: keep the exact location, camera angle, perspective and light of <Picture 1> - <list the visible things>."*
316
+ - Then *"Add <what> into this exact scene"*, and introduce each inserted item from its picture with position, size, facing and depth order.
317
+ - End with how the new parts take the light of Picture 1.
318
+
319
+ ---
320
+
321
+ ## 9. No negative prompt: how to steer away from things
322
+
323
+ The model runs at CFG 1 with no negative prompt, so "negative" control happens only inside the positive prompt.
324
+
325
+ 1. **Describe what should be there instead.** Not "no clutter" but "a bare white desk with only a laptop and a mug". Not "not blurry" but "sharp focus on her eyes".
326
+ 2. **Do not list unwanted objects in detail.** Naming a thing can summon it.
327
+ 3. **A short closing clause is fine for style drift and extras**, because the text encoder is an LLM and understands it: "not illustration, not CGI render look", "no text overlays, no watermark", "no other people". Keep it to one short line at the end.
328
+ 4. **For edits, use "keep" lists** rather than "don't change" lists.
329
+
330
+ ---
331
+
332
+ ## 10. Problems and fixes
333
+
334
+ | Problem | Fix |
335
+ |---|---|
336
+ | Glossy, plastic "doll" face | Longer, physical prompt; name a phone or film camera; "visible skin texture and pores, no retouching" (4.1) |
337
+ | Distant faces broken (teeth, eyes) | Render at 16 MP; or move the person closer; give their figure size |
338
+ | Person looks pasted in | Light direction on the face matches the scene light; cast and contact shadows; same colour temperature as the room |
339
+ | Wrong body direction | Place the target in the frame, then state camera-relative facing (5.1) |
340
+ | Too many people | Exact count; "no other people in the frame"; background people "small, seen from behind" |
341
+ | Reference face drifts or two faces merge | Subject labels (8.2); describe the face in words too; fewer references |
342
+ | Character-sheet look leaks in (grey backdrop, A-pose, game render) | "Use only their faces and outfits. Do not carry over..." (8.2.3) |
343
+ | Object from a reference is huge | Real size in cm and a comparison with a body part |
344
+ | Colour cast (teal, green, orange) | "neutral white balance, true-to-life colours, no colour grading", or "colours match <Picture N> exactly" |
345
+ | Text misspelled | Double quotes, shorter text, larger size, one text item per location |
346
+ | Edit changed more than asked | Longer, more concrete keep list; "Everything else stays exactly the same"; smaller edit |
347
+ | Image looks generic | Replace adjectives with specific objects, materials and wear |
348
+ | Too dark / flat | Name a light source with direction and a fill: "bright window light from screen-left, white wall bouncing soft fill on the right" |
349
+
350
+ ---
351
+
352
+ ## 11. Final check (run before answering)
353
+
354
+ - [ ] Mode is right, and an edit starts with "Edit <Picture 1>:".
355
+ - [ ] Aspect ratio chosen from the request and main subject, not from added detail.
356
+ - [ ] Every request from the user is visible in the prompt.
357
+ - [ ] Exact count of main subjects; adults stated as adults.
358
+ - [ ] Every subject has a place in the frame, a camera-relative facing, and a clear action.
359
+ - [ ] Hands and contact are stated where they matter.
360
+ - [ ] One named light source with direction; shadows agree with it.
361
+ - [ ] Camera: lens, height, angle, focus.
362
+ - [ ] Quoted text is exact, placed, and short.
363
+ - [ ] References: each picture is labelled, described in words, and told what to take and what to leave.
364
+ - [ ] Edits: concrete keep list, change described as a result, light and perspective matched.
365
+ - [ ] No keyword spam; no long list of unwanted things.
366
+ - [ ] Output in the format of section 0, with settings.
367
+
368
+ ---
369
+
370
+ ## 12. Examples
371
+
372
+ ### 12.1 Text-to-image, realistic person
373
+
374
+ **Mode:** text-to-image · **Settings:** 3:4 · draft 4 MP, final 16 MP · Enhance off
375
+
376
+ ```text
377
+ A candid smartphone photograph of an adult woman in her late twenties sitting at a small round marble table outside a Paris café on a grey spring afternoon. She has light olive skin with a few freckles across her nose, dark brown shoulder-length wavy hair with loose flyaway strands, brown eyes, and no visible makeup. She wears an oversized camel wool coat over a white crew-neck T-shirt and thin gold hoop earrings. She sits facing the camera in a front three-quarter view, turned slightly toward screen-left, her right hand wrapped around a white espresso cup on the table and her left elbow resting next to a folded copy of a newspaper. She smiles with her lips closed, looking just past the camera. Behind her, slightly out of focus, wicker café chairs, a green-painted shop front with gold lettering, and two passers-by seen from behind on the wet pavement. Soft overcast daylight from the open sky above and to screen-left, gentle shadows under her chin and on the right side of her face. Shot on the iPhone 16 Pro main camera at eye level from across the table, visible skin texture and pores, natural phone HDR, no retouching, ordinary Instagram photo, not cinematic, not studio.
378
+ ```
379
+
380
+ ### 12.2 Text-to-image, poster with text
381
+
382
+ **Mode:** text-to-image · **Settings:** 2:3 · 16 MP · Enhance off
383
+
384
+ ```text
385
+ A minimalist vintage travel poster in flat screen-print style with a limited palette of deep teal, sand, coral and cream. The upper two thirds show a lone white lighthouse with a red lantern room standing on dark rocks at screen-right, a low coral sun setting over a calm teal sea at screen-left, and three small seagulls drawn as simple cream shapes in the sky. A thin cream border runs around the whole poster. At the top, the headline "CAPE SOLACE" is set in large cream condensed sans-serif capitals with wide letter spacing. In the bottom band of solid sand colour, the line "Where the coast road ends" sits centred in small teal italic serif, and below it the small text "EST. 1874" in coral capitals. Subtle paper grain and slight ink misregistration at the colour edges.
386
+ ```
387
+
388
+ ### 12.3 Edit, one picture
389
+
390
+ **Mode:** edit · **Pictures:** Picture 1 = the user's living room photo · **Settings:** same ratio as Picture 1 · Enhance off
391
+
392
+ ```text
393
+ Edit <Picture 1>: keep the exact room, camera angle, perspective and furniture of <Picture 1> - the grey fabric sofa against the back wall, the wooden coffee table, the tall bookshelf on screen-left and the window on screen-right. The light is now late golden hour: warm low sun enters through the window on screen-right and throws long orange rectangles of light across the floor and the front of the sofa, while the left side of the room falls into soft blue-grey shade. A sleeping orange tabby cat is curled up on the right sofa cushion, about the size of the cushion, lit by the sunbeam, with a soft shadow on the fabric beneath it. Everything else in <Picture 1> stays exactly the same.
394
+ ```
395
+
396
+ ### 12.4 Multi-reference composition
397
+
398
+ **Mode:** multi-reference · **Pictures:** Picture 1 = man (character sheet), Picture 2 = woman (photo), Picture 3 = market street · **Settings:** 3:2 · 16 MP · Enhance off
399
+
400
+ ```text
401
+ <Subject 1> is the adult man in <Picture 1>, a real human with exactly the same face: tanned skin, short black hair with a fade, a trimmed black beard and dark brown eyes. He wears a rust-orange bomber jacket over a black T-shirt, black cargo trousers and white sneakers.
402
+ <Subject 2> is the adult woman in <Picture 2>, with exactly the same face: pale skin, freckles, green eyes and copper-red hair in a loose low bun. She wears a cream knitted cardigan, a sage-green slip dress and brown leather ankle boots.
403
+ <Picture 1> is a character sheet. Use only his face, hair and outfit; do not carry over its grey background, panel layout, A-pose or 3D render look.
404
+ <Picture 3> is the location: a narrow evening street market with fruit stalls under striped red-and-white awnings on the left, strings of warm bulb lights overhead, and a tiled old building with blue shutters at the far end.
405
+
406
+ A candid real-life photograph taken in the street from <Picture 3> just after sunset. <Subject 1> stands on screen-left in a right-facing three-quarter view, holding a paper bag of oranges against his chest with his left arm and offering one orange to <Subject 2> with his right hand. <Subject 2> stands on screen-right in a left-facing three-quarter view, laughing, reaching with her right hand to take the orange; their fingers almost touch in the centre of the frame. A few shoppers in the background are small and seen from behind. Warm light from the bulbs overhead and the stall lamps falls on their faces from above and screen-left, with cool blue dusk sky light filling the shadows. The colours and light match <Picture 3>. Shot handheld at eye level with a 35mm lens, f/2, both of them sharp, the far end of the street softly blurred. Natural skin texture with pores, real fabric, true-to-life colour, nobody looks at the camera.
407
+ ```
408
+
409
+ ---
410
+
411
+ ## 13. Workflow notes (ComfyUI)
412
+
413
+ These match the common ComfyUI setup for this model; other hosts may name things differently.
414
+
415
+ - **Sampler:** 6 steps, CFG 1.0, scheduler simple. LCM-type sampler (s_noise 1.1 → 0.9) gives the most real skin; plain `lcm` at 4 steps or `euler` at 6 also work. Do not raise the steps.
416
+ - **Size:** set width/height on the MiniMax H3 Image Conditioning node; both multiples of 32. Draft at 4 MP, final at 16 MP.
417
+ - **Reference slots:** switch a Load Image slot on (Ctrl+B un-bypasses it), fill from slot 1 upward. Keep the VAE connected to the conditioning node so faces and rooms do not drift.
418
+ - **Enhance prompt (upsampler):** a small LLM expands a one-line idea into a long caption and picks a ratio. Useful for text-to-image when you have no LLM at hand; turn it off for edits and references. If you used this guide, your prompt is already long, so leave it off.
419
+ - **Seed:** fix the seed while tuning wording, so changes come from the prompt, not the noise.
Minimax_H3_image_12gb_vram.json ADDED
@@ -0,0 +1,2108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "id": "840471d0-c494-51f9-b4ba-28a899e23738",
3
+ "revision": 0,
4
+ "last_node_id": 60,
5
+ "last_link_id": 42,
6
+ "nodes": [
7
+ {
8
+ "properties": {
9
+ "Node name for S&R": "UNETLoader"
10
+ },
11
+ "id": 1,
12
+ "type": "UNETLoader",
13
+ "pos": [
14
+ 0,
15
+ 0
16
+ ],
17
+ "size": [
18
+ 420,
19
+ 90
20
+ ],
21
+ "flags": {},
22
+ "order": 0,
23
+ "mode": 0,
24
+ "inputs": [],
25
+ "outputs": [
26
+ {
27
+ "name": "MODEL",
28
+ "type": "MODEL",
29
+ "links": [
30
+ 33,
31
+ 36
32
+ ]
33
+ }
34
+ ],
35
+ "widgets_values": [
36
+ "minimax_h3_image_pruned_int8_convrot.safetensors",
37
+ "default"
38
+ ]
39
+ },
40
+ {
41
+ "properties": {
42
+ "Node name for S&R": "CLIPLoader"
43
+ },
44
+ "id": 2,
45
+ "type": "CLIPLoader",
46
+ "pos": [
47
+ 0,
48
+ 140
49
+ ],
50
+ "size": [
51
+ 420,
52
+ 120
53
+ ],
54
+ "flags": {},
55
+ "order": 1,
56
+ "mode": 0,
57
+ "inputs": [],
58
+ "outputs": [
59
+ {
60
+ "name": "CLIP",
61
+ "type": "CLIP",
62
+ "links": [
63
+ 15
64
+ ]
65
+ }
66
+ ],
67
+ "widgets_values": [
68
+ "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
69
+ "minimax",
70
+ "default"
71
+ ],
72
+ "title": "Load CLIP (text encoder)"
73
+ },
74
+ {
75
+ "properties": {
76
+ "Node name for S&R": "VAELoader"
77
+ },
78
+ "id": 3,
79
+ "type": "VAELoader",
80
+ "pos": [
81
+ 0,
82
+ 310
83
+ ],
84
+ "size": [
85
+ 420,
86
+ 70
87
+ ],
88
+ "flags": {},
89
+ "order": 2,
90
+ "mode": 0,
91
+ "inputs": [],
92
+ "outputs": [
93
+ {
94
+ "name": "VAE",
95
+ "type": "VAE",
96
+ "links": [
97
+ 16,
98
+ 31
99
+ ]
100
+ }
101
+ ],
102
+ "widgets_values": [
103
+ "minimax_h3_image_vae_fp16.safetensors"
104
+ ]
105
+ },
106
+ {
107
+ "properties": {
108
+ "Node name for S&R": "PrimitiveStringMultiline"
109
+ },
110
+ "id": 10,
111
+ "type": "PrimitiveStringMultiline",
112
+ "pos": [
113
+ 0,
114
+ 520
115
+ ],
116
+ "size": [
117
+ 420,
118
+ 330
119
+ ],
120
+ "flags": {},
121
+ "order": 3,
122
+ "mode": 0,
123
+ "inputs": [],
124
+ "outputs": [
125
+ {
126
+ "name": "STRING",
127
+ "type": "STRING",
128
+ "links": [
129
+ 2,
130
+ 7
131
+ ]
132
+ }
133
+ ],
134
+ "widgets_values": [
135
+ "PASTE YOUR PROMPT HERE. With reference pictures, name them as <Picture 1>, <Picture 2>, ..."
136
+ ],
137
+ "title": "PROMPT"
138
+ },
139
+ {
140
+ "properties": {
141
+ "Node name for S&R": "PrimitiveBoolean"
142
+ },
143
+ "id": 11,
144
+ "type": "PrimitiveBoolean",
145
+ "pos": [
146
+ 0,
147
+ 900
148
+ ],
149
+ "size": [
150
+ 420,
151
+ 60
152
+ ],
153
+ "flags": {},
154
+ "order": 4,
155
+ "mode": 0,
156
+ "inputs": [],
157
+ "outputs": [
158
+ {
159
+ "name": "BOOLEAN",
160
+ "type": "BOOLEAN",
161
+ "links": [
162
+ 9
163
+ ]
164
+ }
165
+ ],
166
+ "widgets_values": [
167
+ false
168
+ ],
169
+ "title": "Enhance prompt (text-to-image only)"
170
+ },
171
+ {
172
+ "properties": {
173
+ "Node name for S&R": "CLIPLoader"
174
+ },
175
+ "id": 12,
176
+ "type": "CLIPLoader",
177
+ "pos": [
178
+ 0,
179
+ 1010
180
+ ],
181
+ "size": [
182
+ 420,
183
+ 120
184
+ ],
185
+ "flags": {},
186
+ "order": 5,
187
+ "mode": 0,
188
+ "inputs": [],
189
+ "outputs": [
190
+ {
191
+ "name": "CLIP",
192
+ "type": "CLIP",
193
+ "links": [
194
+ 3
195
+ ]
196
+ }
197
+ ],
198
+ "widgets_values": [
199
+ "qwen3.5_9b_w4a8_gptq.safetensors",
200
+ "minimax",
201
+ "default"
202
+ ],
203
+ "title": "Load CLIP (prompt upsampler)"
204
+ },
205
+ {
206
+ "properties": {
207
+ "Node name for S&R": "PrimitiveStringMultiline"
208
+ },
209
+ "id": 13,
210
+ "type": "PrimitiveStringMultiline",
211
+ "pos": [
212
+ 0,
213
+ 1180
214
+ ],
215
+ "size": [
216
+ 420,
217
+ 120
218
+ ],
219
+ "flags": {},
220
+ "order": 6,
221
+ "mode": 0,
222
+ "inputs": [],
223
+ "outputs": [
224
+ {
225
+ "name": "STRING",
226
+ "type": "STRING",
227
+ "links": [
228
+ 4
229
+ ]
230
+ }
231
+ ],
232
+ "widgets_values": [
233
+ "Reason in the thinking section first. The visible final content must be exactly one JSON object with exactly two keys: expanded_caption and aspect_ratio. Every ASCII double quotation mark used as caption content inside expanded_caption must be serialized as the JSON escape sequence \\\". Do not output text outside that object."
234
+ ],
235
+ "title": "Upsampler system prompt"
236
+ },
237
+ {
238
+ "properties": {
239
+ "Node name for S&R": "PrimitiveStringMultiline"
240
+ },
241
+ "id": 14,
242
+ "type": "PrimitiveStringMultiline",
243
+ "pos": [
244
+ 0,
245
+ 1350
246
+ ],
247
+ "size": [
248
+ 420,
249
+ 120
250
+ ],
251
+ "flags": {},
252
+ "order": 7,
253
+ "mode": 0,
254
+ "inputs": [],
255
+ "outputs": [
256
+ {
257
+ "name": "STRING",
258
+ "type": "STRING",
259
+ "links": [
260
+ 1
261
+ ]
262
+ }
263
+ ],
264
+ "widgets_values": [
265
+ "You are a caption-expansion and canvas-selection model for still-image\ntext-to-image generation. Rewrite input_text into a natural English detailed\ncaption and choose one aspect ratio for the requested final image. After private\nreasoning, serialize the complete caption and aspect ratio together as the one\nrequired two-key JSON object; no visible analysis, markdown, or self-talk. The\ncaption may be long when the requested image is a complex information or design artifact.\n\nRole: act like an expert visual designer and technical illustrator, not a word\nexpander. Preserve every hard fact from the input, then complete the image with\nvisible, inspectable content. The goal is not more words; the goal is that a\nviewer can inspect each important part of the image and find meaningful subjects,\nlabels, relationships, materials, data, and hierarchy.\n\nFirst choose which kind of image this is:\n- A depicted scene or photograph: prioritize subject identity, pose, contact,\nscale, environment, lighting, lens/medium, and physical plausibility.\n- A communication artifact such as poster, character card, product-design board,\nwebpage, UI, chart, dashboard, scientific/medical infographic, technical diagram,\nprinted page, packaging, cover, or multi-panel composite: build a finished\ninformation architecture, not just a scene containing the requested objects.\n\nAspect-ratio decision policy:\nChoose exactly one of these allowed ratios: 1:1, 5:4, 4:3, 3:2, 16:9, 21:9,\n4:5, 3:4, 2:3, or 9:16. Decide in the following order; earlier evidence\noverrides later evidence.\n\n1. Explicit final canvas\n- Preserve an explicit final-output ratio. Map explicit output pixel dimensions\n by width divided by height to the closest allowed ratio.\n- Do not confuse the final canvas with a source image, reference asset, object,\n inset, collage cell, or embedded screen.\n\n2. Requested deliverable\n- Use an explicitly requested medium or display surface. Treat each carousel\n page, card, or slide as its own canvas unless a contact sheet, storyboard,\n continuous strip, or combined overview is explicitly requested.\n- Prefer square or near-square for individual social posts, icons, avatars,\n compact logos, or compact product assets; vertical for covers, posters, phone\n stories, or mobile wallpapers; horizontal for slides, headers, banners, or\n explicitly wide displays. Do not invent a platform or deliverable.\n\n3. Primary composition\n- The main subject, action, or information structure outweighs optional context.\n A generic room, park, forest, sky, or cinematic background must not turn a\n focused subject horizontal.\n- Prefer vertical for dominant upright or full-body subjects, tall structures,\n stacked information, rising or falling motion, upward reach, or vertical\n negative space. Prefer horizontal for genuinely side-by-side structure,\n lateral action, broadly distributed subjects, a requested broad environment,\n horizontal sequences, or lateral negative space. Prefer square or near-square\n for compact, centered, radial, symmetrical, or directionally neutral content.\n\n4. Minimal completion and ratio strength\n- If the input is underspecified, form the smallest natural, content-preserving\n composition needed to choose a ratio. Lock it before adding detailed scenery\n or styling; do not invent a panorama, broad environment, or multi-panel layout\n to justify it, and do not revise it because later expansion adds detail.\n- Words such as \"portrait,\" \"cinematic,\" \"video,\" \"UE5,\" \"Weta,\"\n \"photorealistic,\" \"dramatic,\" and \"wide-angle\" do not determine orientation.\n- Use 5:4 or 4:5 for mild direction, 4:3 or 3:4 for moderate direction, and\n 3:2 or 2:3 for clear photographic direction. Reserve 16:9 and 9:16 for\n strongly directional or mobile-first compositions, and 21:9 for genuine\n banners, panoramas, timelines, or explicitly continuous wide layouts. When\n evidence remains weak, prefer 1:1 or a moderate ratio.\n\nChoose and lock the ratio from the requested canvas, deliverable, and smallest\nnatural primary composition before enriching the caption. Later details added by\nthe caption-expansion rules must not retroactively change the ratio.\n\nUniversal hard rules:\n1. Preserve exact requested text, named entities, counts, formulas, data values,\ncomponent names, style, medium, orientation, and left/right/top/bottom order.\n2. Required content must be visible, not implied. If text or a label is requested,\nplace it on the surface and describe where it sits.\n3. For active bodies, solve scene geometry before pose detail. Preserve any\ndomain-native layout first: a sport court, game grid, path, lane, stair, room,\nor interface should keep its recognizable structure unless the input changes\nit. Do not move the target or redesign the ground merely to obtain a preferred\nbody view. Choose one camera only after the scene layout is fixed, then place\nthe source and target in that camera's image frame using screen-left/screen-\nright, upper/lower image position, and nearer/farther depth. Silently trace the\nsource-to-target vector in this fixed view. Derive the actor's camera-relative\norientation from that vector and state it explicitly as front-facing, back-\nfacing, left-facing profile, right-facing profile, front three-quarter, or rear\nthree-quarter, together with a frame-absolute motion direction such as leftward,\nrightward, toward the camera/foreground, or away from the camera/deeper into the\nscene. Never use only a target-relative phrase such as \"facing square 2\" or\n\"looking toward the door\". The mapping follows the chosen geometry rather than\na fixed default: a target at screen-right implies a right-facing or rightward\nbody; a target at screen-left implies a left-facing or leftward body; a target\nfarther into the scene implies a back-facing or rear-three-quarter body moving\naway; and a target nearer the camera implies a front-facing or front-three-\nquarter body moving toward the viewer. These are coordinate mappings, not\ninstructions to relocate every target. Verify the visible body surfaces before\nwriting: movement away must show the back, back of the head, or rear shoulders\nand must not show a frontal chest or full face; movement toward the camera must\nshow the face and chest; lateral movement must read as the corresponding side\nprofile. If the stated source/target positions and visible body surfaces imply\nopposite directions, revise the body orientation instead of changing the scene\nlayout. Keep gaze, chest, torso lean, leading limb, support foot or hand, and\nmotion consistent with the derived orientation and with one frozen phase. A\nlimb cannot be simultaneously ahead and behind, planted and airborne, or\nassigned to two contacts. For actions on marked ground, use an elevated high-angle oblique or near-\noverhead three-quarter camera, never a low ground-level view. Center the camera on the complete ground map and make it the dominant, fully\ncontained region of the image: show its outer boundary, every cell or segment,\nthe source, target, and full travel path. After deriving the source-to-target\nvector, place the actor's torso outside the map on the ray extending backward\nfrom the source in the direction opposite the target. Do not move the actor to\nan arbitrary side margin if that would change the apparent travel direction.\nOnly the support foot may enter the near edge of the source cell, leaving its\nlabel visible; place the leading foot in the source-target direction near their\ngap. Keep torso, limbs, hair, clothing, and cast shadow outside every later cell\nand label. Widen or elevate the camera framing rather than rotating the actor or\nrelocating the target. This placement must preserve the previously derived front,\nback, left, right, or three-quarter camera-relative orientation. Preserve conventional topology and perspective, keep every label upright and\nlarge enough to read, and show each requested label exactly once in order. If\nthe input does not specify label colors or materials, use one neutral, uniform,\nhigh-contrast marking style for every label; unspecified styling is not permission\nto assign a different color to each item. Keep the actor's hands empty unless the\ninput names a held object. Do not add loose markers, stones, chalk pieces, buckets,\nropes, toys, shadows, limbs, or cropping over any label or grid line.\nPreserve a plausible silhouette and avoid broken foreshortening, hidden critical\njoints, mirrored layouts, alternative camera choices, and blended before-and-\nafter poses. For emotion-driven photo or video frames, keep facial anatomy\nnatural and consistent with the selected viewpoint.\n4. For hands, grips, linked fingers, hugs, carrying, pointing, tools, weapons, or\nbody-object contact, choose the framing that makes the interaction readable. Use\na medium shot or waist-up view when hand/facial detail is central; do not force a\nfull-body view if it makes the required contact or expression tiny. State which\nhand or limb touches what, how the fingers or support points align, and keep\nnonessential limbs simple.\n5. Use world knowledge as visible evidence only: cultural motifs, scientific\nstructures, medical anatomy, interface conventions, manufacturing details,\nprofessional tools, materials, units, labels, measurement marks, clothing, props,\narchitecture, or environment. Do not add lore that cannot be seen.\n6. Respect requested subject count. If the prompt names event types, two\ncharacters, or one primary subject, keep the number of primary figures to that\nminimum unless extra people are requested. Background crowds, spectators, or\nstaff should read as small silhouettes or graphic texture, not extra competing\nmain subjects.\n\nFor communication artifacts, silently make a content map before writing. Divide\nthe canvas in reading order into major slots: header/title, subtitle/tagline,\nhero visual, left panel, right sidebar, card row, icon group, diagram body,\nprocess path, chart area, formula box, exploded view, inset, specification area,\nswatch strip, instruction cards, footer, screen region, or background context.\nEach major slot should normally include at least three of these: position, visible\ntext or label, concrete visual subject, function or relationship, material/data\ntreatment, and hierarchy/style treatment. Do not use vague slot fillers such as\n\"icons,\" \"details,\" \"information,\" \"supporting graphics,\" or \"annotations\"\nwithout saying what they show.\n\nMake the artifact feel complete. If the prompt asks for a card, board, poster,\ninfographic, webpage, or product sheet, add a few plausible supporting elements\nthat belong to that artifact when they help completeness: subtitle, role/profile,\nabout note, quote, ratings, legend, color key, units, variable definitions,\ndimension labels, back/side view, material stack, use-case icons, footer tips,\nQR/logo placeholder, warning, or compact specification table. Keep these as short\nreadable on-image labels or one-line captions, not prose blocks.\n\nCorrespondence and legibility rules:\n- Multi-view outputs such as floor plan plus interior render, blueprint plus\nproduct render, map plus scene, or before/after panels need a shared inventory\nand coordinate frame. Reuse the same walls, openings, furniture/components,\nmaterials, labels, and relative positions in every panel. Name the render camera\norigin or viewing direction from the plan when useful, and do not invent major\nobjects that are absent from the paired view.\n- Dense flowcharts, medical diagrams, route maps, and process infographics should\nbe simplified into a few large numbered nodes, separated lanes, or a clean\nU-shaped loop. Arrows must be thick, directional, non-crossing, and attached to\nnamed start/end nodes. Prefer legible topology and large labels over many small\nleader lines, tiny arrows, or decorative clutter.\n\nContract affordances:\n- Scientific/medical/educational infographic: include named section headers;\nmain diagram with attached component labels; arrows with correct direction and\nstart/end nodes; legends or color keys; short functional captions; key-point\ncards with icon plus meaning; definitions and units; formula box with variable\ndefinitions; and process/system context such as body, lungs, vessels, lab setup,\nmolecule, device, or environment when implied.\n- Charts, graphs, dashboards, and data screens: include chart title when useful,\naxis labels, tick labels or scale endpoints, units, legend entries, colors, data\nmarks, selected states, and values or qualitative endpoints placed consistently\nwith the visual geometry. Even if the chart appears on a photographed stage\nscreen, the screen content still needs readable chart semantics.\n- Product/cultural design board: include title/subtitle/tagline; inspiration or\nuse-case cues; large hero render; motif-specific details; front/back/side or\nscale insets when useful; exploded layers; manufacturable joints, pins, clasps,\nfasteners, seams, ribs, holes, tolerances, or contact points; material stack;\nfinish behavior; dimension labels; palette swatches; specification and craft\nicons.\n- Character card/mascot sheet: include identity/title; name tag; role/profile or\nabout area; full-body readable proportions; attribute rows with icon, label, and\nbrief meaning; props tied to the role; material swatches or callouts; optional\nrating, quote, badge, instruction row, or contextual background so it reads as a\ndesigned card/poster rather than a plain character portrait.\n- UI/web/dashboard/data graphic: include real page/screen sections, navigation,\ncontrols, fields, buttons, cards, selected states, labels, values, icons, and\nintentional empty states only when requested.\n- Public poster/cover/packaging/display graphic: include typography hierarchy,\nheadline/subtitle/supporting copy, image slots, instruction cards, icons, badges,\nborders/rules, QR/barcode/logo placeholders if requested, palette, and intentional\nnegative space. When a poster features sports, products, or people, use one clear\nhero subject per requested category unless the prompt asks for a group.\n\nStyle comes after content. Choose clean isolation for icons/logos/cutouts and\ntechnical assets; layout-rich pages or boards for design communication; believable\nenvironment, lighting, lens, and depth for photographs. Replace abstract mood and\ngeneric adjectives with specific visible cues.\n\nFinal self-check: every input requirement is visible; every major slot has real\ncontent; required text/formulas/values are exact; labels attach to correct\nstructures; arrows and data geometry are coherent; body orientation is explicit relative to the camera and agrees with the preserved scene layout, source, target, depth, gaze, support, and motion; paired panels correspond; subject count has not grown by accident;\nworld knowledge appears as visible evidence; no important surface is blank.\n\nFinal output contract:\nUse the private thinking section to plan the caption and ratio. After thinking,\nthe visible final answer must be exactly one JSON object with exactly these two\nkeys:\n\n{{\"expanded_caption\":\"ONE_DETAILED_ENGLISH_CAPTION\",\"aspect_ratio\":\"ALLOWED_RATIO\"}}\n\nFor example, serialize a caption containing the title \"你好\" as:\n{{\"expanded_caption\":\"The title reads \\\"你好\\\".\",\"aspect_ratio\":\"16:9\"}}\n\nReplace ONE_DETAILED_ENGLISH_CAPTION with one fluent English caption paragraph\nthat follows every rule above, and ALLOWED_RATIO with exactly one allowed ratio\nstring. Keep expanded_caption as one valid JSON string. Every ASCII double quotation mark used as caption content inside expanded_caption must be serialized as the JSON escape sequence \\\". Escape all other characters as valid JSON requires. Do not add or omit keys. Do not put thinking or think tags inside\nthe object. Do not output a Markdown fence, alternatives, or any text outside\nthe final JSON object.\n\ninput_text: {input text}\noutput_text:\n"
266
+ ],
267
+ "title": "Upsampler task prompt"
268
+ },
269
+ {
270
+ "properties": {
271
+ "Node name for S&R": "StringReplace"
272
+ },
273
+ "id": 15,
274
+ "type": "StringReplace",
275
+ "pos": [
276
+ 470,
277
+ 900
278
+ ],
279
+ "size": [
280
+ 300,
281
+ 200
282
+ ],
283
+ "flags": {},
284
+ "order": 8,
285
+ "mode": 0,
286
+ "inputs": [
287
+ {
288
+ "name": "string",
289
+ "type": "STRING",
290
+ "link": 1,
291
+ "widget": {
292
+ "name": "string"
293
+ }
294
+ },
295
+ {
296
+ "name": "find",
297
+ "type": "STRING",
298
+ "link": null,
299
+ "widget": {
300
+ "name": "find"
301
+ }
302
+ },
303
+ {
304
+ "name": "replace",
305
+ "type": "STRING",
306
+ "link": 2,
307
+ "widget": {
308
+ "name": "replace"
309
+ }
310
+ }
311
+ ],
312
+ "outputs": [
313
+ {
314
+ "name": "STRING",
315
+ "type": "STRING",
316
+ "links": [
317
+ 5
318
+ ]
319
+ }
320
+ ],
321
+ "widgets_values": [
322
+ "",
323
+ "{input text}",
324
+ ""
325
+ ]
326
+ },
327
+ {
328
+ "properties": {
329
+ "Node name for S&R": "TextGenerateV2"
330
+ },
331
+ "id": 16,
332
+ "type": "TextGenerateV2",
333
+ "pos": [
334
+ 470,
335
+ 1160
336
+ ],
337
+ "size": [
338
+ 300,
339
+ 460
340
+ ],
341
+ "flags": {},
342
+ "order": 9,
343
+ "mode": 0,
344
+ "inputs": [
345
+ {
346
+ "name": "clip",
347
+ "type": "CLIP",
348
+ "link": 3
349
+ },
350
+ {
351
+ "name": "system_prompt",
352
+ "type": "STRING",
353
+ "link": 4
354
+ },
355
+ {
356
+ "name": "prompt",
357
+ "type": "STRING",
358
+ "link": 5,
359
+ "widget": {
360
+ "name": "prompt"
361
+ }
362
+ }
363
+ ],
364
+ "outputs": [
365
+ {
366
+ "name": "generated_text",
367
+ "type": "STRING",
368
+ "links": [
369
+ 6
370
+ ]
371
+ },
372
+ {
373
+ "name": "thinking",
374
+ "type": "STRING",
375
+ "links": []
376
+ }
377
+ ],
378
+ "widgets_values": [
379
+ "",
380
+ 8192,
381
+ "on",
382
+ 1,
383
+ 20,
384
+ 0.95,
385
+ 0,
386
+ 1,
387
+ 0,
388
+ 0,
389
+ true,
390
+ true,
391
+ "auto"
392
+ ]
393
+ },
394
+ {
395
+ "properties": {
396
+ "Node name for S&R": "JsonExtractString"
397
+ },
398
+ "id": 17,
399
+ "type": "JsonExtractString",
400
+ "pos": [
401
+ 0,
402
+ 1520
403
+ ],
404
+ "size": [
405
+ 420,
406
+ 110
407
+ ],
408
+ "flags": {},
409
+ "order": 10,
410
+ "mode": 0,
411
+ "inputs": [
412
+ {
413
+ "name": "json_string",
414
+ "type": "STRING",
415
+ "link": 6,
416
+ "widget": {
417
+ "name": "json_string"
418
+ }
419
+ },
420
+ {
421
+ "name": "key",
422
+ "type": "STRING",
423
+ "link": null,
424
+ "widget": {
425
+ "name": "key"
426
+ }
427
+ }
428
+ ],
429
+ "outputs": [
430
+ {
431
+ "name": "STRING",
432
+ "type": "STRING",
433
+ "links": [
434
+ 8
435
+ ]
436
+ }
437
+ ],
438
+ "widgets_values": [
439
+ "",
440
+ "expanded_caption"
441
+ ]
442
+ },
443
+ {
444
+ "properties": {
445
+ "Node name for S&R": "ComfySwitchNode"
446
+ },
447
+ "id": 18,
448
+ "type": "ComfySwitchNode",
449
+ "pos": [
450
+ 470,
451
+ 520
452
+ ],
453
+ "size": [
454
+ 300,
455
+ 90
456
+ ],
457
+ "flags": {},
458
+ "order": 11,
459
+ "mode": 0,
460
+ "inputs": [
461
+ {
462
+ "name": "on_false",
463
+ "type": "STRING",
464
+ "link": 7
465
+ },
466
+ {
467
+ "name": "on_true",
468
+ "type": "STRING",
469
+ "link": 8
470
+ },
471
+ {
472
+ "name": "switch",
473
+ "type": "BOOLEAN",
474
+ "link": 9,
475
+ "widget": {
476
+ "name": "switch"
477
+ }
478
+ }
479
+ ],
480
+ "outputs": [
481
+ {
482
+ "name": "output",
483
+ "type": "STRING",
484
+ "links": [
485
+ 10,
486
+ 27
487
+ ]
488
+ }
489
+ ],
490
+ "widgets_values": [
491
+ false
492
+ ],
493
+ "title": "Prompt: as written / enhanced"
494
+ },
495
+ {
496
+ "properties": {
497
+ "Node name for S&R": "PreviewAny"
498
+ },
499
+ "id": 19,
500
+ "type": "PreviewAny",
501
+ "pos": [
502
+ 470,
503
+ 660
504
+ ],
505
+ "size": [
506
+ 300,
507
+ 190
508
+ ],
509
+ "flags": {},
510
+ "order": 12,
511
+ "mode": 0,
512
+ "inputs": [
513
+ {
514
+ "name": "source",
515
+ "type": "*",
516
+ "link": 10
517
+ }
518
+ ],
519
+ "outputs": [
520
+ {
521
+ "name": "STRING",
522
+ "type": "STRING",
523
+ "links": []
524
+ }
525
+ ],
526
+ "widgets_values": [],
527
+ "title": "Prompt sent to the model"
528
+ },
529
+ {
530
+ "properties": {
531
+ "Node name for S&R": "PrimitiveBoolean"
532
+ },
533
+ "id": 20,
534
+ "type": "PrimitiveBoolean",
535
+ "pos": [
536
+ 840,
537
+ 0
538
+ ],
539
+ "size": [
540
+ 300,
541
+ 60
542
+ ],
543
+ "flags": {},
544
+ "order": 13,
545
+ "mode": 0,
546
+ "inputs": [],
547
+ "outputs": [
548
+ {
549
+ "name": "BOOLEAN",
550
+ "type": "BOOLEAN",
551
+ "links": [
552
+ 13
553
+ ]
554
+ }
555
+ ],
556
+ "widgets_values": [
557
+ true
558
+ ],
559
+ "title": "16 MP (off = 4 MP draft)"
560
+ },
561
+ {
562
+ "properties": {
563
+ "Node name for S&R": "PrimitiveFloat"
564
+ },
565
+ "id": 21,
566
+ "type": "PrimitiveFloat",
567
+ "pos": [
568
+ 840,
569
+ 110
570
+ ],
571
+ "size": [
572
+ 300,
573
+ 60
574
+ ],
575
+ "flags": {},
576
+ "order": 14,
577
+ "mode": 0,
578
+ "inputs": [],
579
+ "outputs": [
580
+ {
581
+ "name": "FLOAT",
582
+ "type": "FLOAT",
583
+ "links": [
584
+ 11
585
+ ]
586
+ }
587
+ ],
588
+ "widgets_values": [
589
+ 4.0
590
+ ],
591
+ "title": "Draft megapixels"
592
+ },
593
+ {
594
+ "properties": {
595
+ "Node name for S&R": "PrimitiveFloat"
596
+ },
597
+ "id": 22,
598
+ "type": "PrimitiveFloat",
599
+ "pos": [
600
+ 840,
601
+ 220
602
+ ],
603
+ "size": [
604
+ 300,
605
+ 60
606
+ ],
607
+ "flags": {},
608
+ "order": 15,
609
+ "mode": 0,
610
+ "inputs": [],
611
+ "outputs": [
612
+ {
613
+ "name": "FLOAT",
614
+ "type": "FLOAT",
615
+ "links": [
616
+ 12
617
+ ]
618
+ }
619
+ ],
620
+ "widgets_values": [
621
+ 16.0
622
+ ],
623
+ "title": "Full megapixels (max 16)"
624
+ },
625
+ {
626
+ "properties": {
627
+ "Node name for S&R": "ComfySwitchNode"
628
+ },
629
+ "id": 23,
630
+ "type": "ComfySwitchNode",
631
+ "pos": [
632
+ 840,
633
+ 330
634
+ ],
635
+ "size": [
636
+ 300,
637
+ 90
638
+ ],
639
+ "flags": {},
640
+ "order": 16,
641
+ "mode": 0,
642
+ "inputs": [
643
+ {
644
+ "name": "on_false",
645
+ "type": "FLOAT",
646
+ "link": 11
647
+ },
648
+ {
649
+ "name": "on_true",
650
+ "type": "FLOAT",
651
+ "link": 12
652
+ },
653
+ {
654
+ "name": "switch",
655
+ "type": "BOOLEAN",
656
+ "link": 13,
657
+ "widget": {
658
+ "name": "switch"
659
+ }
660
+ }
661
+ ],
662
+ "outputs": [
663
+ {
664
+ "name": "output",
665
+ "type": "FLOAT",
666
+ "links": [
667
+ 14
668
+ ]
669
+ }
670
+ ],
671
+ "widgets_values": [
672
+ false
673
+ ],
674
+ "title": "Megapixels: draft / 16 MP"
675
+ },
676
+ {
677
+ "properties": {
678
+ "Node name for S&R": "ResolutionSelector"
679
+ },
680
+ "id": 24,
681
+ "type": "ResolutionSelector",
682
+ "pos": [
683
+ 1190,
684
+ 0
685
+ ],
686
+ "size": [
687
+ 300,
688
+ 240
689
+ ],
690
+ "flags": {},
691
+ "order": 17,
692
+ "mode": 0,
693
+ "inputs": [
694
+ {
695
+ "name": "aspect_ratio",
696
+ "type": "COMBO",
697
+ "link": null,
698
+ "widget": {
699
+ "name": "aspect_ratio"
700
+ }
701
+ },
702
+ {
703
+ "name": "megapixels",
704
+ "type": "FLOAT",
705
+ "link": 14,
706
+ "widget": {
707
+ "name": "megapixels"
708
+ }
709
+ },
710
+ {
711
+ "name": "multiple",
712
+ "type": "INT",
713
+ "link": null,
714
+ "widget": {
715
+ "name": "multiple"
716
+ }
717
+ }
718
+ ],
719
+ "outputs": [
720
+ {
721
+ "name": "width",
722
+ "type": "INT",
723
+ "links": [
724
+ 28
725
+ ]
726
+ },
727
+ {
728
+ "name": "height",
729
+ "type": "INT",
730
+ "links": [
731
+ 29
732
+ ]
733
+ }
734
+ ],
735
+ "widgets_values": [
736
+ "16:9 (Widescreen)",
737
+ 16.0,
738
+ 32
739
+ ],
740
+ "title": "ASPECT RATIO"
741
+ },
742
+ {
743
+ "properties": {
744
+ "Node name for S&R": "MiniMaxH3ImageConditioning"
745
+ },
746
+ "id": 30,
747
+ "type": "MiniMaxH3ImageConditioning",
748
+ "pos": [
749
+ 1560,
750
+ 0
751
+ ],
752
+ "size": [
753
+ 330,
754
+ 400
755
+ ],
756
+ "flags": {},
757
+ "order": 18,
758
+ "mode": 0,
759
+ "inputs": [
760
+ {
761
+ "name": "clip",
762
+ "type": "CLIP",
763
+ "link": 15
764
+ },
765
+ {
766
+ "name": "vae",
767
+ "type": "VAE",
768
+ "link": 16
769
+ },
770
+ {
771
+ "name": "images.image_1",
772
+ "type": "IMAGE",
773
+ "link": 17
774
+ },
775
+ {
776
+ "name": "images.image_2",
777
+ "type": "IMAGE",
778
+ "link": 18
779
+ },
780
+ {
781
+ "name": "images.image_3",
782
+ "type": "IMAGE",
783
+ "link": 19
784
+ },
785
+ {
786
+ "name": "images.image_4",
787
+ "type": "IMAGE",
788
+ "link": 20
789
+ },
790
+ {
791
+ "name": "images.image_5",
792
+ "type": "IMAGE",
793
+ "link": 21
794
+ },
795
+ {
796
+ "name": "images.image_6",
797
+ "type": "IMAGE",
798
+ "link": 22
799
+ },
800
+ {
801
+ "name": "images.image_7",
802
+ "type": "IMAGE",
803
+ "link": 23
804
+ },
805
+ {
806
+ "name": "images.image_8",
807
+ "type": "IMAGE",
808
+ "link": 24
809
+ },
810
+ {
811
+ "name": "images.image_9",
812
+ "type": "IMAGE",
813
+ "link": 25
814
+ },
815
+ {
816
+ "name": "images.image_10",
817
+ "type": "IMAGE",
818
+ "link": 26
819
+ },
820
+ {
821
+ "name": "prompt",
822
+ "type": "STRING",
823
+ "link": 27,
824
+ "widget": {
825
+ "name": "prompt"
826
+ }
827
+ },
828
+ {
829
+ "name": "width",
830
+ "type": "INT",
831
+ "link": 28,
832
+ "widget": {
833
+ "name": "width"
834
+ }
835
+ },
836
+ {
837
+ "name": "height",
838
+ "type": "INT",
839
+ "link": 29,
840
+ "widget": {
841
+ "name": "height"
842
+ }
843
+ }
844
+ ],
845
+ "outputs": [
846
+ {
847
+ "name": "positive",
848
+ "type": "CONDITIONING",
849
+ "links": [
850
+ 30,
851
+ 34
852
+ ]
853
+ },
854
+ {
855
+ "name": "LATENT",
856
+ "type": "LATENT",
857
+ "links": [
858
+ 41
859
+ ]
860
+ }
861
+ ],
862
+ "widgets_values": [
863
+ "",
864
+ 0,
865
+ 0
866
+ ]
867
+ },
868
+ {
869
+ "properties": {
870
+ "Node name for S&R": "ConditioningZeroOut"
871
+ },
872
+ "id": 31,
873
+ "type": "ConditioningZeroOut",
874
+ "pos": [
875
+ 1940,
876
+ 0
877
+ ],
878
+ "size": [
879
+ 300,
880
+ 50
881
+ ],
882
+ "flags": {},
883
+ "order": 19,
884
+ "mode": 0,
885
+ "inputs": [
886
+ {
887
+ "name": "conditioning",
888
+ "type": "CONDITIONING",
889
+ "link": 30
890
+ }
891
+ ],
892
+ "outputs": [
893
+ {
894
+ "name": "CONDITIONING",
895
+ "type": "CONDITIONING",
896
+ "links": [
897
+ 35
898
+ ]
899
+ }
900
+ ]
901
+ },
902
+ {
903
+ "properties": {
904
+ "Node name for S&R": "VAEDecodeTiled"
905
+ },
906
+ "id": 33,
907
+ "type": "VAEDecodeTiled",
908
+ "pos": [
909
+ 2290,
910
+ 0
911
+ ],
912
+ "size": [
913
+ 260,
914
+ 150
915
+ ],
916
+ "flags": {},
917
+ "order": 20,
918
+ "mode": 0,
919
+ "inputs": [
920
+ {
921
+ "name": "samples",
922
+ "type": "LATENT",
923
+ "link": 42
924
+ },
925
+ {
926
+ "name": "vae",
927
+ "type": "VAE",
928
+ "link": 31
929
+ },
930
+ {
931
+ "name": "tile_size",
932
+ "type": "INT",
933
+ "link": null,
934
+ "widget": {
935
+ "name": "tile_size"
936
+ }
937
+ },
938
+ {
939
+ "name": "overlap",
940
+ "type": "INT",
941
+ "link": null,
942
+ "widget": {
943
+ "name": "overlap"
944
+ }
945
+ },
946
+ {
947
+ "name": "temporal_size",
948
+ "type": "INT",
949
+ "link": null,
950
+ "widget": {
951
+ "name": "temporal_size"
952
+ }
953
+ },
954
+ {
955
+ "name": "temporal_overlap",
956
+ "type": "INT",
957
+ "link": null,
958
+ "widget": {
959
+ "name": "temporal_overlap"
960
+ }
961
+ }
962
+ ],
963
+ "outputs": [
964
+ {
965
+ "name": "IMAGE",
966
+ "type": "IMAGE",
967
+ "links": [
968
+ 32
969
+ ]
970
+ }
971
+ ],
972
+ "title": "VAE Decode (tiled, low VRAM)",
973
+ "widgets_values": [
974
+ 768,
975
+ 64,
976
+ 64,
977
+ 8
978
+ ]
979
+ },
980
+ {
981
+ "properties": {
982
+ "Node name for S&R": "SaveImage"
983
+ },
984
+ "id": 34,
985
+ "type": "SaveImage",
986
+ "pos": [
987
+ 2290,
988
+ 200
989
+ ],
990
+ "size": [
991
+ 560,
992
+ 520
993
+ ],
994
+ "flags": {},
995
+ "order": 21,
996
+ "mode": 0,
997
+ "inputs": [
998
+ {
999
+ "name": "images",
1000
+ "type": "IMAGE",
1001
+ "link": 32
1002
+ }
1003
+ ],
1004
+ "outputs": [],
1005
+ "widgets_values": [
1006
+ "h3_image/minimax_h3_image_12gb"
1007
+ ]
1008
+ },
1009
+ {
1010
+ "properties": {
1011
+ "Node name for S&R": "RandomNoise"
1012
+ },
1013
+ "id": 35,
1014
+ "type": "RandomNoise",
1015
+ "pos": [
1016
+ 1940,
1017
+ 90
1018
+ ],
1019
+ "size": [
1020
+ 300,
1021
+ 82
1022
+ ],
1023
+ "flags": {},
1024
+ "order": 22,
1025
+ "mode": 0,
1026
+ "inputs": [
1027
+ {
1028
+ "name": "noise_seed",
1029
+ "type": "INT",
1030
+ "link": null,
1031
+ "widget": {
1032
+ "name": "noise_seed"
1033
+ }
1034
+ }
1035
+ ],
1036
+ "outputs": [
1037
+ {
1038
+ "name": "NOISE",
1039
+ "type": "NOISE",
1040
+ "links": [
1041
+ 37
1042
+ ]
1043
+ }
1044
+ ],
1045
+ "widgets_values": [
1046
+ 42,
1047
+ "randomize"
1048
+ ],
1049
+ "title": "Seed"
1050
+ },
1051
+ {
1052
+ "properties": {
1053
+ "Node name for S&R": "CFGGuider"
1054
+ },
1055
+ "id": 36,
1056
+ "type": "CFGGuider",
1057
+ "pos": [
1058
+ 1940,
1059
+ 200
1060
+ ],
1061
+ "size": [
1062
+ 300,
1063
+ 100
1064
+ ],
1065
+ "flags": {},
1066
+ "order": 23,
1067
+ "mode": 0,
1068
+ "inputs": [
1069
+ {
1070
+ "name": "model",
1071
+ "type": "MODEL",
1072
+ "link": 33
1073
+ },
1074
+ {
1075
+ "name": "positive",
1076
+ "type": "CONDITIONING",
1077
+ "link": 34
1078
+ },
1079
+ {
1080
+ "name": "negative",
1081
+ "type": "CONDITIONING",
1082
+ "link": 35
1083
+ },
1084
+ {
1085
+ "name": "cfg",
1086
+ "type": "FLOAT",
1087
+ "link": null,
1088
+ "widget": {
1089
+ "name": "cfg"
1090
+ }
1091
+ }
1092
+ ],
1093
+ "outputs": [
1094
+ {
1095
+ "name": "GUIDER",
1096
+ "type": "GUIDER",
1097
+ "links": [
1098
+ 38
1099
+ ]
1100
+ }
1101
+ ],
1102
+ "widgets_values": [
1103
+ 1.0
1104
+ ]
1105
+ },
1106
+ {
1107
+ "properties": {
1108
+ "Node name for S&R": "BasicScheduler"
1109
+ },
1110
+ "id": 37,
1111
+ "type": "BasicScheduler",
1112
+ "pos": [
1113
+ 1940,
1114
+ 330
1115
+ ],
1116
+ "size": [
1117
+ 300,
1118
+ 110
1119
+ ],
1120
+ "flags": {},
1121
+ "order": 24,
1122
+ "mode": 0,
1123
+ "inputs": [
1124
+ {
1125
+ "name": "model",
1126
+ "type": "MODEL",
1127
+ "link": 36
1128
+ },
1129
+ {
1130
+ "name": "scheduler",
1131
+ "type": "COMBO",
1132
+ "link": null,
1133
+ "widget": {
1134
+ "name": "scheduler"
1135
+ }
1136
+ },
1137
+ {
1138
+ "name": "steps",
1139
+ "type": "INT",
1140
+ "link": null,
1141
+ "widget": {
1142
+ "name": "steps"
1143
+ }
1144
+ },
1145
+ {
1146
+ "name": "denoise",
1147
+ "type": "FLOAT",
1148
+ "link": null,
1149
+ "widget": {
1150
+ "name": "denoise"
1151
+ }
1152
+ }
1153
+ ],
1154
+ "outputs": [
1155
+ {
1156
+ "name": "SIGMAS",
1157
+ "type": "SIGMAS",
1158
+ "links": [
1159
+ 40
1160
+ ]
1161
+ }
1162
+ ],
1163
+ "widgets_values": [
1164
+ "simple",
1165
+ 6,
1166
+ 1.0
1167
+ ]
1168
+ },
1169
+ {
1170
+ "properties": {
1171
+ "Node name for S&R": "SamplerLCM"
1172
+ },
1173
+ "id": 38,
1174
+ "type": "SamplerLCM",
1175
+ "pos": [
1176
+ 1940,
1177
+ 470
1178
+ ],
1179
+ "size": [
1180
+ 300,
1181
+ 110
1182
+ ],
1183
+ "flags": {},
1184
+ "order": 25,
1185
+ "mode": 0,
1186
+ "inputs": [
1187
+ {
1188
+ "name": "s_noise",
1189
+ "type": "FLOAT",
1190
+ "link": null,
1191
+ "widget": {
1192
+ "name": "s_noise"
1193
+ }
1194
+ },
1195
+ {
1196
+ "name": "s_noise_end",
1197
+ "type": "FLOAT",
1198
+ "link": null,
1199
+ "widget": {
1200
+ "name": "s_noise_end"
1201
+ }
1202
+ },
1203
+ {
1204
+ "name": "noise_clip_std",
1205
+ "type": "FLOAT",
1206
+ "link": null,
1207
+ "widget": {
1208
+ "name": "noise_clip_std"
1209
+ }
1210
+ }
1211
+ ],
1212
+ "outputs": [
1213
+ {
1214
+ "name": "SAMPLER",
1215
+ "type": "SAMPLER",
1216
+ "links": [
1217
+ 39
1218
+ ]
1219
+ }
1220
+ ],
1221
+ "widgets_values": [
1222
+ 1.1,
1223
+ 0.9,
1224
+ 0.0
1225
+ ],
1226
+ "title": "SamplerLCM (real skin)"
1227
+ },
1228
+ {
1229
+ "properties": {
1230
+ "Node name for S&R": "SamplerCustomAdvanced"
1231
+ },
1232
+ "id": 39,
1233
+ "type": "SamplerCustomAdvanced",
1234
+ "pos": [
1235
+ 1560,
1236
+ 440
1237
+ ],
1238
+ "size": [
1239
+ 330,
1240
+ 130
1241
+ ],
1242
+ "flags": {},
1243
+ "order": 26,
1244
+ "mode": 0,
1245
+ "inputs": [
1246
+ {
1247
+ "name": "noise",
1248
+ "type": "NOISE",
1249
+ "link": 37
1250
+ },
1251
+ {
1252
+ "name": "guider",
1253
+ "type": "GUIDER",
1254
+ "link": 38
1255
+ },
1256
+ {
1257
+ "name": "sampler",
1258
+ "type": "SAMPLER",
1259
+ "link": 39
1260
+ },
1261
+ {
1262
+ "name": "sigmas",
1263
+ "type": "SIGMAS",
1264
+ "link": 40
1265
+ },
1266
+ {
1267
+ "name": "latent_image",
1268
+ "type": "LATENT",
1269
+ "link": 41
1270
+ }
1271
+ ],
1272
+ "outputs": [
1273
+ {
1274
+ "name": "output",
1275
+ "type": "LATENT",
1276
+ "links": [
1277
+ 42
1278
+ ]
1279
+ },
1280
+ {
1281
+ "name": "denoised_output",
1282
+ "type": "LATENT",
1283
+ "links": []
1284
+ }
1285
+ ]
1286
+ },
1287
+ {
1288
+ "properties": {
1289
+ "Node name for S&R": "LoadImage"
1290
+ },
1291
+ "id": 41,
1292
+ "type": "LoadImage",
1293
+ "pos": [
1294
+ 840,
1295
+ 790
1296
+ ],
1297
+ "size": [
1298
+ 270,
1299
+ 360
1300
+ ],
1301
+ "flags": {},
1302
+ "order": 27,
1303
+ "mode": 4,
1304
+ "inputs": [],
1305
+ "outputs": [
1306
+ {
1307
+ "name": "IMAGE",
1308
+ "type": "IMAGE",
1309
+ "links": [
1310
+ 17
1311
+ ]
1312
+ },
1313
+ {
1314
+ "name": "MASK",
1315
+ "type": "MASK",
1316
+ "links": []
1317
+ }
1318
+ ],
1319
+ "widgets_values": [
1320
+ "01_Elena.png",
1321
+ "image"
1322
+ ],
1323
+ "title": "Picture 1 (the one to edit)"
1324
+ },
1325
+ {
1326
+ "properties": {
1327
+ "Node name for S&R": "LoadImage"
1328
+ },
1329
+ "id": 42,
1330
+ "type": "LoadImage",
1331
+ "pos": [
1332
+ 1130,
1333
+ 790
1334
+ ],
1335
+ "size": [
1336
+ 270,
1337
+ 360
1338
+ ],
1339
+ "flags": {},
1340
+ "order": 28,
1341
+ "mode": 4,
1342
+ "inputs": [],
1343
+ "outputs": [
1344
+ {
1345
+ "name": "IMAGE",
1346
+ "type": "IMAGE",
1347
+ "links": [
1348
+ 18
1349
+ ]
1350
+ },
1351
+ {
1352
+ "name": "MASK",
1353
+ "type": "MASK",
1354
+ "links": []
1355
+ }
1356
+ ],
1357
+ "widgets_values": [
1358
+ "01_Elena.png",
1359
+ "image"
1360
+ ],
1361
+ "title": "Picture 2"
1362
+ },
1363
+ {
1364
+ "properties": {
1365
+ "Node name for S&R": "LoadImage"
1366
+ },
1367
+ "id": 43,
1368
+ "type": "LoadImage",
1369
+ "pos": [
1370
+ 1420,
1371
+ 790
1372
+ ],
1373
+ "size": [
1374
+ 270,
1375
+ 360
1376
+ ],
1377
+ "flags": {},
1378
+ "order": 29,
1379
+ "mode": 4,
1380
+ "inputs": [],
1381
+ "outputs": [
1382
+ {
1383
+ "name": "IMAGE",
1384
+ "type": "IMAGE",
1385
+ "links": [
1386
+ 19
1387
+ ]
1388
+ },
1389
+ {
1390
+ "name": "MASK",
1391
+ "type": "MASK",
1392
+ "links": []
1393
+ }
1394
+ ],
1395
+ "widgets_values": [
1396
+ "01_Elena.png",
1397
+ "image"
1398
+ ],
1399
+ "title": "Picture 3"
1400
+ },
1401
+ {
1402
+ "properties": {
1403
+ "Node name for S&R": "LoadImage"
1404
+ },
1405
+ "id": 44,
1406
+ "type": "LoadImage",
1407
+ "pos": [
1408
+ 1710,
1409
+ 790
1410
+ ],
1411
+ "size": [
1412
+ 270,
1413
+ 360
1414
+ ],
1415
+ "flags": {},
1416
+ "order": 30,
1417
+ "mode": 4,
1418
+ "inputs": [],
1419
+ "outputs": [
1420
+ {
1421
+ "name": "IMAGE",
1422
+ "type": "IMAGE",
1423
+ "links": [
1424
+ 20
1425
+ ]
1426
+ },
1427
+ {
1428
+ "name": "MASK",
1429
+ "type": "MASK",
1430
+ "links": []
1431
+ }
1432
+ ],
1433
+ "widgets_values": [
1434
+ "01_Elena.png",
1435
+ "image"
1436
+ ],
1437
+ "title": "Picture 4"
1438
+ },
1439
+ {
1440
+ "properties": {
1441
+ "Node name for S&R": "LoadImage"
1442
+ },
1443
+ "id": 45,
1444
+ "type": "LoadImage",
1445
+ "pos": [
1446
+ 2000,
1447
+ 790
1448
+ ],
1449
+ "size": [
1450
+ 270,
1451
+ 360
1452
+ ],
1453
+ "flags": {},
1454
+ "order": 31,
1455
+ "mode": 4,
1456
+ "inputs": [],
1457
+ "outputs": [
1458
+ {
1459
+ "name": "IMAGE",
1460
+ "type": "IMAGE",
1461
+ "links": [
1462
+ 21
1463
+ ]
1464
+ },
1465
+ {
1466
+ "name": "MASK",
1467
+ "type": "MASK",
1468
+ "links": []
1469
+ }
1470
+ ],
1471
+ "widgets_values": [
1472
+ "01_Elena.png",
1473
+ "image"
1474
+ ],
1475
+ "title": "Picture 5"
1476
+ },
1477
+ {
1478
+ "properties": {
1479
+ "Node name for S&R": "LoadImage"
1480
+ },
1481
+ "id": 46,
1482
+ "type": "LoadImage",
1483
+ "pos": [
1484
+ 840,
1485
+ 1210
1486
+ ],
1487
+ "size": [
1488
+ 270,
1489
+ 360
1490
+ ],
1491
+ "flags": {},
1492
+ "order": 32,
1493
+ "mode": 4,
1494
+ "inputs": [],
1495
+ "outputs": [
1496
+ {
1497
+ "name": "IMAGE",
1498
+ "type": "IMAGE",
1499
+ "links": [
1500
+ 22
1501
+ ]
1502
+ },
1503
+ {
1504
+ "name": "MASK",
1505
+ "type": "MASK",
1506
+ "links": []
1507
+ }
1508
+ ],
1509
+ "widgets_values": [
1510
+ "01_Elena.png",
1511
+ "image"
1512
+ ],
1513
+ "title": "Picture 6"
1514
+ },
1515
+ {
1516
+ "properties": {
1517
+ "Node name for S&R": "LoadImage"
1518
+ },
1519
+ "id": 47,
1520
+ "type": "LoadImage",
1521
+ "pos": [
1522
+ 1130,
1523
+ 1210
1524
+ ],
1525
+ "size": [
1526
+ 270,
1527
+ 360
1528
+ ],
1529
+ "flags": {},
1530
+ "order": 33,
1531
+ "mode": 4,
1532
+ "inputs": [],
1533
+ "outputs": [
1534
+ {
1535
+ "name": "IMAGE",
1536
+ "type": "IMAGE",
1537
+ "links": [
1538
+ 23
1539
+ ]
1540
+ },
1541
+ {
1542
+ "name": "MASK",
1543
+ "type": "MASK",
1544
+ "links": []
1545
+ }
1546
+ ],
1547
+ "widgets_values": [
1548
+ "01_Elena.png",
1549
+ "image"
1550
+ ],
1551
+ "title": "Picture 7"
1552
+ },
1553
+ {
1554
+ "properties": {
1555
+ "Node name for S&R": "LoadImage"
1556
+ },
1557
+ "id": 48,
1558
+ "type": "LoadImage",
1559
+ "pos": [
1560
+ 1420,
1561
+ 1210
1562
+ ],
1563
+ "size": [
1564
+ 270,
1565
+ 360
1566
+ ],
1567
+ "flags": {},
1568
+ "order": 34,
1569
+ "mode": 4,
1570
+ "inputs": [],
1571
+ "outputs": [
1572
+ {
1573
+ "name": "IMAGE",
1574
+ "type": "IMAGE",
1575
+ "links": [
1576
+ 24
1577
+ ]
1578
+ },
1579
+ {
1580
+ "name": "MASK",
1581
+ "type": "MASK",
1582
+ "links": []
1583
+ }
1584
+ ],
1585
+ "widgets_values": [
1586
+ "01_Elena.png",
1587
+ "image"
1588
+ ],
1589
+ "title": "Picture 8"
1590
+ },
1591
+ {
1592
+ "properties": {
1593
+ "Node name for S&R": "LoadImage"
1594
+ },
1595
+ "id": 49,
1596
+ "type": "LoadImage",
1597
+ "pos": [
1598
+ 1710,
1599
+ 1210
1600
+ ],
1601
+ "size": [
1602
+ 270,
1603
+ 360
1604
+ ],
1605
+ "flags": {},
1606
+ "order": 35,
1607
+ "mode": 4,
1608
+ "inputs": [],
1609
+ "outputs": [
1610
+ {
1611
+ "name": "IMAGE",
1612
+ "type": "IMAGE",
1613
+ "links": [
1614
+ 25
1615
+ ]
1616
+ },
1617
+ {
1618
+ "name": "MASK",
1619
+ "type": "MASK",
1620
+ "links": []
1621
+ }
1622
+ ],
1623
+ "widgets_values": [
1624
+ "01_Elena.png",
1625
+ "image"
1626
+ ],
1627
+ "title": "Picture 9"
1628
+ },
1629
+ {
1630
+ "properties": {
1631
+ "Node name for S&R": "LoadImage"
1632
+ },
1633
+ "id": 50,
1634
+ "type": "LoadImage",
1635
+ "pos": [
1636
+ 2000,
1637
+ 1210
1638
+ ],
1639
+ "size": [
1640
+ 270,
1641
+ 360
1642
+ ],
1643
+ "flags": {},
1644
+ "order": 36,
1645
+ "mode": 4,
1646
+ "inputs": [],
1647
+ "outputs": [
1648
+ {
1649
+ "name": "IMAGE",
1650
+ "type": "IMAGE",
1651
+ "links": [
1652
+ 26
1653
+ ]
1654
+ },
1655
+ {
1656
+ "name": "MASK",
1657
+ "type": "MASK",
1658
+ "links": []
1659
+ }
1660
+ ],
1661
+ "widgets_values": [
1662
+ "01_Elena.png",
1663
+ "image"
1664
+ ],
1665
+ "title": "Picture 10"
1666
+ },
1667
+ {
1668
+ "properties": {
1669
+ "Node name for S&R": "MarkdownNote"
1670
+ },
1671
+ "id": 60,
1672
+ "type": "MarkdownNote",
1673
+ "pos": [
1674
+ -570,
1675
+ 0
1676
+ ],
1677
+ "size": [
1678
+ 520,
1679
+ 1100
1680
+ ],
1681
+ "flags": {},
1682
+ "order": 37,
1683
+ "mode": 0,
1684
+ "inputs": [],
1685
+ "outputs": [],
1686
+ "widgets_values": [
1687
+ "## MiniMax H3 Image - 16 MP on 12 GB VRAM\n\nSame model, sampler and **16 MP** final size as the 16 GB version. ComfyUI streams the model through the card in parts, so a smaller card gives the **same image**, only a little slower. The VAE decode is tiled to keep the last step small.\n\n**1. Prompt** - write it **long**, like a real photo: who (age, hair, skin, outfit), where, what is around them, the light and its direction, and the camera, e.g. *\"Shot on the iPhone 16 Pro front camera, visible skin texture and pores, natural phone HDR, no retouching, ordinary Instagram photo, not cinematic, not studio.\"* Short prompts give a glossy doll face. *Enhance prompt* (Qwen3.5 upsampler) can write the long version for text-to-image; leave it off with reference pictures.\n\n**2. Size** - pick the ASPECT RATIO. *16 MP* on = final, off = 4 MP draft for testing prompts fast. People: use 3:4 (or 4:3).\n\n| Ratio | draft 4 MP | 16 MP |\n|---|---|---|\n| 1:1 | 2048x2048 | 4096x4096 |\n| 3:4 | 1760x2368 | 3552x4736 |\n| 4:3 | 2368x1760 | 4736x3552 |\n| 2:3 | 1664x2496 | 3360x5024 |\n| 3:2 | 2496x1664 | 5024x3360 |\n| 9:16 | 1536x2720 | 3072x5472 |\n| 16:9 | 2720x1536 | 5472x3072 |\n| 21:9 | 3136x1344 | 6272x2688 |\n\n**3. Reference pictures** - 10 slots, all bypassed. Select a slot, press **Ctrl+B** to switch it on, and load your picture. Fill slots from 1 upward: with slots 1, 2 and 3 on they are `<Picture 1>`, `<Picture 2>`, `<Picture 3>` in the prompt. Picture 1 is the one an edit instruction changes. Keep the VAE connected to the conditioning node so the face and room do not drift.\n\n**4. Sampler** - SamplerLCM (s_noise 1.1 -> 0.9), BasicScheduler **simple, 6 steps**, CFGGuider **cfg 1.0**. Do not raise the steps: the model is distilled for 6.\n\n**Decode** - VAE Decode (tiled), tile 768. If the decode runs out of memory, lower tile_size to 384 or 256.\n\n**Measured on 12 GB** (RTX 5070 Ti limited to the VRAM a 12 GB card has free): text-to-image at **16 MP in about 1 minute** (4.8 s per step), the identical image to the 16 GB version with the same seed. A real 12 GB card may have a slower chip.\n\n**12 GB tips**\n- Use the **4 MP draft** while you work on the prompt, then *16 MP* for the final.\n- Up to 10 reference pictures fit, but **4 or fewer** keeps it fast.\n- System RAM: tested with 24 GB.\n"
1688
+ ],
1689
+ "title": "How to use (12 GB VRAM)"
1690
+ }
1691
+ ],
1692
+ "links": [
1693
+ [
1694
+ 1,
1695
+ 14,
1696
+ 0,
1697
+ 15,
1698
+ 0,
1699
+ "STRING"
1700
+ ],
1701
+ [
1702
+ 2,
1703
+ 10,
1704
+ 0,
1705
+ 15,
1706
+ 2,
1707
+ "STRING"
1708
+ ],
1709
+ [
1710
+ 3,
1711
+ 12,
1712
+ 0,
1713
+ 16,
1714
+ 0,
1715
+ "CLIP"
1716
+ ],
1717
+ [
1718
+ 4,
1719
+ 13,
1720
+ 0,
1721
+ 16,
1722
+ 1,
1723
+ "STRING"
1724
+ ],
1725
+ [
1726
+ 5,
1727
+ 15,
1728
+ 0,
1729
+ 16,
1730
+ 2,
1731
+ "STRING"
1732
+ ],
1733
+ [
1734
+ 6,
1735
+ 16,
1736
+ 0,
1737
+ 17,
1738
+ 0,
1739
+ "STRING"
1740
+ ],
1741
+ [
1742
+ 7,
1743
+ 10,
1744
+ 0,
1745
+ 18,
1746
+ 0,
1747
+ "STRING"
1748
+ ],
1749
+ [
1750
+ 8,
1751
+ 17,
1752
+ 0,
1753
+ 18,
1754
+ 1,
1755
+ "STRING"
1756
+ ],
1757
+ [
1758
+ 9,
1759
+ 11,
1760
+ 0,
1761
+ 18,
1762
+ 2,
1763
+ "BOOLEAN"
1764
+ ],
1765
+ [
1766
+ 10,
1767
+ 18,
1768
+ 0,
1769
+ 19,
1770
+ 0,
1771
+ "STRING"
1772
+ ],
1773
+ [
1774
+ 11,
1775
+ 21,
1776
+ 0,
1777
+ 23,
1778
+ 0,
1779
+ "FLOAT"
1780
+ ],
1781
+ [
1782
+ 12,
1783
+ 22,
1784
+ 0,
1785
+ 23,
1786
+ 1,
1787
+ "FLOAT"
1788
+ ],
1789
+ [
1790
+ 13,
1791
+ 20,
1792
+ 0,
1793
+ 23,
1794
+ 2,
1795
+ "BOOLEAN"
1796
+ ],
1797
+ [
1798
+ 14,
1799
+ 23,
1800
+ 0,
1801
+ 24,
1802
+ 1,
1803
+ "FLOAT"
1804
+ ],
1805
+ [
1806
+ 15,
1807
+ 2,
1808
+ 0,
1809
+ 30,
1810
+ 0,
1811
+ "CLIP"
1812
+ ],
1813
+ [
1814
+ 16,
1815
+ 3,
1816
+ 0,
1817
+ 30,
1818
+ 1,
1819
+ "VAE"
1820
+ ],
1821
+ [
1822
+ 17,
1823
+ 41,
1824
+ 0,
1825
+ 30,
1826
+ 2,
1827
+ "IMAGE"
1828
+ ],
1829
+ [
1830
+ 18,
1831
+ 42,
1832
+ 0,
1833
+ 30,
1834
+ 3,
1835
+ "IMAGE"
1836
+ ],
1837
+ [
1838
+ 19,
1839
+ 43,
1840
+ 0,
1841
+ 30,
1842
+ 4,
1843
+ "IMAGE"
1844
+ ],
1845
+ [
1846
+ 20,
1847
+ 44,
1848
+ 0,
1849
+ 30,
1850
+ 5,
1851
+ "IMAGE"
1852
+ ],
1853
+ [
1854
+ 21,
1855
+ 45,
1856
+ 0,
1857
+ 30,
1858
+ 6,
1859
+ "IMAGE"
1860
+ ],
1861
+ [
1862
+ 22,
1863
+ 46,
1864
+ 0,
1865
+ 30,
1866
+ 7,
1867
+ "IMAGE"
1868
+ ],
1869
+ [
1870
+ 23,
1871
+ 47,
1872
+ 0,
1873
+ 30,
1874
+ 8,
1875
+ "IMAGE"
1876
+ ],
1877
+ [
1878
+ 24,
1879
+ 48,
1880
+ 0,
1881
+ 30,
1882
+ 9,
1883
+ "IMAGE"
1884
+ ],
1885
+ [
1886
+ 25,
1887
+ 49,
1888
+ 0,
1889
+ 30,
1890
+ 10,
1891
+ "IMAGE"
1892
+ ],
1893
+ [
1894
+ 26,
1895
+ 50,
1896
+ 0,
1897
+ 30,
1898
+ 11,
1899
+ "IMAGE"
1900
+ ],
1901
+ [
1902
+ 27,
1903
+ 18,
1904
+ 0,
1905
+ 30,
1906
+ 12,
1907
+ "STRING"
1908
+ ],
1909
+ [
1910
+ 28,
1911
+ 24,
1912
+ 0,
1913
+ 30,
1914
+ 13,
1915
+ "INT"
1916
+ ],
1917
+ [
1918
+ 29,
1919
+ 24,
1920
+ 1,
1921
+ 30,
1922
+ 14,
1923
+ "INT"
1924
+ ],
1925
+ [
1926
+ 30,
1927
+ 30,
1928
+ 0,
1929
+ 31,
1930
+ 0,
1931
+ "CONDITIONING"
1932
+ ],
1933
+ [
1934
+ 31,
1935
+ 3,
1936
+ 0,
1937
+ 33,
1938
+ 1,
1939
+ "VAE"
1940
+ ],
1941
+ [
1942
+ 32,
1943
+ 33,
1944
+ 0,
1945
+ 34,
1946
+ 0,
1947
+ "IMAGE"
1948
+ ],
1949
+ [
1950
+ 33,
1951
+ 1,
1952
+ 0,
1953
+ 36,
1954
+ 0,
1955
+ "MODEL"
1956
+ ],
1957
+ [
1958
+ 34,
1959
+ 30,
1960
+ 0,
1961
+ 36,
1962
+ 1,
1963
+ "CONDITIONING"
1964
+ ],
1965
+ [
1966
+ 35,
1967
+ 31,
1968
+ 0,
1969
+ 36,
1970
+ 2,
1971
+ "CONDITIONING"
1972
+ ],
1973
+ [
1974
+ 36,
1975
+ 1,
1976
+ 0,
1977
+ 37,
1978
+ 0,
1979
+ "MODEL"
1980
+ ],
1981
+ [
1982
+ 37,
1983
+ 35,
1984
+ 0,
1985
+ 39,
1986
+ 0,
1987
+ "NOISE"
1988
+ ],
1989
+ [
1990
+ 38,
1991
+ 36,
1992
+ 0,
1993
+ 39,
1994
+ 1,
1995
+ "GUIDER"
1996
+ ],
1997
+ [
1998
+ 39,
1999
+ 38,
2000
+ 0,
2001
+ 39,
2002
+ 2,
2003
+ "SAMPLER"
2004
+ ],
2005
+ [
2006
+ 40,
2007
+ 37,
2008
+ 0,
2009
+ 39,
2010
+ 3,
2011
+ "SIGMAS"
2012
+ ],
2013
+ [
2014
+ 41,
2015
+ 30,
2016
+ 1,
2017
+ 39,
2018
+ 4,
2019
+ "LATENT"
2020
+ ],
2021
+ [
2022
+ 42,
2023
+ 39,
2024
+ 0,
2025
+ 33,
2026
+ 0,
2027
+ "LATENT"
2028
+ ]
2029
+ ],
2030
+ "groups": [
2031
+ {
2032
+ "id": 1,
2033
+ "title": "Models",
2034
+ "bounding": [
2035
+ -20,
2036
+ -80,
2037
+ 460,
2038
+ 480
2039
+ ],
2040
+ "color": "#3f789e",
2041
+ "font_size": 24,
2042
+ "flags": {}
2043
+ },
2044
+ {
2045
+ "id": 2,
2046
+ "title": "Prompt",
2047
+ "bounding": [
2048
+ -20,
2049
+ 440,
2050
+ 810,
2051
+ 1240
2052
+ ],
2053
+ "color": "#a1309b",
2054
+ "font_size": 24,
2055
+ "flags": {}
2056
+ },
2057
+ {
2058
+ "id": 3,
2059
+ "title": "Size: aspect ratio + 16 MP (12 GB VRAM)",
2060
+ "bounding": [
2061
+ 820,
2062
+ -80,
2063
+ 690,
2064
+ 520
2065
+ ],
2066
+ "color": "#88A",
2067
+ "font_size": 24,
2068
+ "flags": {}
2069
+ },
2070
+ {
2071
+ "id": 4,
2072
+ "title": "Reference pictures - 4 or fewer is fastest (Ctrl+B to switch a slot on)",
2073
+ "bounding": [
2074
+ 820,
2075
+ 710,
2076
+ 1470,
2077
+ 870
2078
+ ],
2079
+ "color": "#8A8",
2080
+ "font_size": 24,
2081
+ "flags": {}
2082
+ },
2083
+ {
2084
+ "id": 5,
2085
+ "title": "Render",
2086
+ "bounding": [
2087
+ 1540,
2088
+ -80,
2089
+ 1330,
2090
+ 750
2091
+ ],
2092
+ "color": "#b58b2a",
2093
+ "font_size": 24,
2094
+ "flags": {}
2095
+ }
2096
+ ],
2097
+ "config": {},
2098
+ "extra": {
2099
+ "ds": {
2100
+ "scale": 0.5,
2101
+ "offset": [
2102
+ 600,
2103
+ 150
2104
+ ]
2105
+ }
2106
+ },
2107
+ "version": 0.4
2108
+ }
Minimax_H3_image_16gb_vram.json ADDED
@@ -0,0 +1,2069 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "id": "3c1398ea-a2bf-5e45-935d-e1f286eb5000",
3
+ "revision": 0,
4
+ "last_node_id": 60,
5
+ "last_link_id": 42,
6
+ "nodes": [
7
+ {
8
+ "properties": {
9
+ "Node name for S&R": "UNETLoader"
10
+ },
11
+ "id": 1,
12
+ "type": "UNETLoader",
13
+ "pos": [
14
+ 0,
15
+ 0
16
+ ],
17
+ "size": [
18
+ 420,
19
+ 90
20
+ ],
21
+ "flags": {},
22
+ "order": 0,
23
+ "mode": 0,
24
+ "inputs": [],
25
+ "outputs": [
26
+ {
27
+ "name": "MODEL",
28
+ "type": "MODEL",
29
+ "links": [
30
+ 33,
31
+ 36
32
+ ]
33
+ }
34
+ ],
35
+ "widgets_values": [
36
+ "minimax_h3_image_pruned_int8_convrot.safetensors",
37
+ "default"
38
+ ]
39
+ },
40
+ {
41
+ "properties": {
42
+ "Node name for S&R": "CLIPLoader"
43
+ },
44
+ "id": 2,
45
+ "type": "CLIPLoader",
46
+ "pos": [
47
+ 0,
48
+ 140
49
+ ],
50
+ "size": [
51
+ 420,
52
+ 120
53
+ ],
54
+ "flags": {},
55
+ "order": 1,
56
+ "mode": 0,
57
+ "inputs": [],
58
+ "outputs": [
59
+ {
60
+ "name": "CLIP",
61
+ "type": "CLIP",
62
+ "links": [
63
+ 15
64
+ ]
65
+ }
66
+ ],
67
+ "widgets_values": [
68
+ "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
69
+ "minimax",
70
+ "default"
71
+ ],
72
+ "title": "Load CLIP (text encoder)"
73
+ },
74
+ {
75
+ "properties": {
76
+ "Node name for S&R": "VAELoader"
77
+ },
78
+ "id": 3,
79
+ "type": "VAELoader",
80
+ "pos": [
81
+ 0,
82
+ 310
83
+ ],
84
+ "size": [
85
+ 420,
86
+ 70
87
+ ],
88
+ "flags": {},
89
+ "order": 2,
90
+ "mode": 0,
91
+ "inputs": [],
92
+ "outputs": [
93
+ {
94
+ "name": "VAE",
95
+ "type": "VAE",
96
+ "links": [
97
+ 16,
98
+ 31
99
+ ]
100
+ }
101
+ ],
102
+ "widgets_values": [
103
+ "minimax_h3_image_vae_fp16.safetensors"
104
+ ]
105
+ },
106
+ {
107
+ "properties": {
108
+ "Node name for S&R": "PrimitiveStringMultiline"
109
+ },
110
+ "id": 10,
111
+ "type": "PrimitiveStringMultiline",
112
+ "pos": [
113
+ 0,
114
+ 520
115
+ ],
116
+ "size": [
117
+ 420,
118
+ 330
119
+ ],
120
+ "flags": {},
121
+ "order": 3,
122
+ "mode": 0,
123
+ "inputs": [],
124
+ "outputs": [
125
+ {
126
+ "name": "STRING",
127
+ "type": "STRING",
128
+ "links": [
129
+ 2,
130
+ 7
131
+ ]
132
+ }
133
+ ],
134
+ "widgets_values": [
135
+ "PASTE YOUR PROMPT HERE. With reference pictures, name them as <Picture 1>, <Picture 2>, ..."
136
+ ],
137
+ "title": "PROMPT"
138
+ },
139
+ {
140
+ "properties": {
141
+ "Node name for S&R": "PrimitiveBoolean"
142
+ },
143
+ "id": 11,
144
+ "type": "PrimitiveBoolean",
145
+ "pos": [
146
+ 0,
147
+ 900
148
+ ],
149
+ "size": [
150
+ 420,
151
+ 60
152
+ ],
153
+ "flags": {},
154
+ "order": 4,
155
+ "mode": 0,
156
+ "inputs": [],
157
+ "outputs": [
158
+ {
159
+ "name": "BOOLEAN",
160
+ "type": "BOOLEAN",
161
+ "links": [
162
+ 9
163
+ ]
164
+ }
165
+ ],
166
+ "widgets_values": [
167
+ false
168
+ ],
169
+ "title": "Enhance prompt (text-to-image only)"
170
+ },
171
+ {
172
+ "properties": {
173
+ "Node name for S&R": "CLIPLoader"
174
+ },
175
+ "id": 12,
176
+ "type": "CLIPLoader",
177
+ "pos": [
178
+ 0,
179
+ 1010
180
+ ],
181
+ "size": [
182
+ 420,
183
+ 120
184
+ ],
185
+ "flags": {},
186
+ "order": 5,
187
+ "mode": 0,
188
+ "inputs": [],
189
+ "outputs": [
190
+ {
191
+ "name": "CLIP",
192
+ "type": "CLIP",
193
+ "links": [
194
+ 3
195
+ ]
196
+ }
197
+ ],
198
+ "widgets_values": [
199
+ "qwen3.5_9b_w4a8_gptq.safetensors",
200
+ "minimax",
201
+ "default"
202
+ ],
203
+ "title": "Load CLIP (prompt upsampler)"
204
+ },
205
+ {
206
+ "properties": {
207
+ "Node name for S&R": "PrimitiveStringMultiline"
208
+ },
209
+ "id": 13,
210
+ "type": "PrimitiveStringMultiline",
211
+ "pos": [
212
+ 0,
213
+ 1180
214
+ ],
215
+ "size": [
216
+ 420,
217
+ 120
218
+ ],
219
+ "flags": {},
220
+ "order": 6,
221
+ "mode": 0,
222
+ "inputs": [],
223
+ "outputs": [
224
+ {
225
+ "name": "STRING",
226
+ "type": "STRING",
227
+ "links": [
228
+ 4
229
+ ]
230
+ }
231
+ ],
232
+ "widgets_values": [
233
+ "Reason in the thinking section first. The visible final content must be exactly one JSON object with exactly two keys: expanded_caption and aspect_ratio. Every ASCII double quotation mark used as caption content inside expanded_caption must be serialized as the JSON escape sequence \\\". Do not output text outside that object."
234
+ ],
235
+ "title": "Upsampler system prompt"
236
+ },
237
+ {
238
+ "properties": {
239
+ "Node name for S&R": "PrimitiveStringMultiline"
240
+ },
241
+ "id": 14,
242
+ "type": "PrimitiveStringMultiline",
243
+ "pos": [
244
+ 0,
245
+ 1350
246
+ ],
247
+ "size": [
248
+ 420,
249
+ 120
250
+ ],
251
+ "flags": {},
252
+ "order": 7,
253
+ "mode": 0,
254
+ "inputs": [],
255
+ "outputs": [
256
+ {
257
+ "name": "STRING",
258
+ "type": "STRING",
259
+ "links": [
260
+ 1
261
+ ]
262
+ }
263
+ ],
264
+ "widgets_values": [
265
+ "You are a caption-expansion and canvas-selection model for still-image\ntext-to-image generation. Rewrite input_text into a natural English detailed\ncaption and choose one aspect ratio for the requested final image. After private\nreasoning, serialize the complete caption and aspect ratio together as the one\nrequired two-key JSON object; no visible analysis, markdown, or self-talk. The\ncaption may be long when the requested image is a complex information or design artifact.\n\nRole: act like an expert visual designer and technical illustrator, not a word\nexpander. Preserve every hard fact from the input, then complete the image with\nvisible, inspectable content. The goal is not more words; the goal is that a\nviewer can inspect each important part of the image and find meaningful subjects,\nlabels, relationships, materials, data, and hierarchy.\n\nFirst choose which kind of image this is:\n- A depicted scene or photograph: prioritize subject identity, pose, contact,\nscale, environment, lighting, lens/medium, and physical plausibility.\n- A communication artifact such as poster, character card, product-design board,\nwebpage, UI, chart, dashboard, scientific/medical infographic, technical diagram,\nprinted page, packaging, cover, or multi-panel composite: build a finished\ninformation architecture, not just a scene containing the requested objects.\n\nAspect-ratio decision policy:\nChoose exactly one of these allowed ratios: 1:1, 5:4, 4:3, 3:2, 16:9, 21:9,\n4:5, 3:4, 2:3, or 9:16. Decide in the following order; earlier evidence\noverrides later evidence.\n\n1. Explicit final canvas\n- Preserve an explicit final-output ratio. Map explicit output pixel dimensions\n by width divided by height to the closest allowed ratio.\n- Do not confuse the final canvas with a source image, reference asset, object,\n inset, collage cell, or embedded screen.\n\n2. Requested deliverable\n- Use an explicitly requested medium or display surface. Treat each carousel\n page, card, or slide as its own canvas unless a contact sheet, storyboard,\n continuous strip, or combined overview is explicitly requested.\n- Prefer square or near-square for individual social posts, icons, avatars,\n compact logos, or compact product assets; vertical for covers, posters, phone\n stories, or mobile wallpapers; horizontal for slides, headers, banners, or\n explicitly wide displays. Do not invent a platform or deliverable.\n\n3. Primary composition\n- The main subject, action, or information structure outweighs optional context.\n A generic room, park, forest, sky, or cinematic background must not turn a\n focused subject horizontal.\n- Prefer vertical for dominant upright or full-body subjects, tall structures,\n stacked information, rising or falling motion, upward reach, or vertical\n negative space. Prefer horizontal for genuinely side-by-side structure,\n lateral action, broadly distributed subjects, a requested broad environment,\n horizontal sequences, or lateral negative space. Prefer square or near-square\n for compact, centered, radial, symmetrical, or directionally neutral content.\n\n4. Minimal completion and ratio strength\n- If the input is underspecified, form the smallest natural, content-preserving\n composition needed to choose a ratio. Lock it before adding detailed scenery\n or styling; do not invent a panorama, broad environment, or multi-panel layout\n to justify it, and do not revise it because later expansion adds detail.\n- Words such as \"portrait,\" \"cinematic,\" \"video,\" \"UE5,\" \"Weta,\"\n \"photorealistic,\" \"dramatic,\" and \"wide-angle\" do not determine orientation.\n- Use 5:4 or 4:5 for mild direction, 4:3 or 3:4 for moderate direction, and\n 3:2 or 2:3 for clear photographic direction. Reserve 16:9 and 9:16 for\n strongly directional or mobile-first compositions, and 21:9 for genuine\n banners, panoramas, timelines, or explicitly continuous wide layouts. When\n evidence remains weak, prefer 1:1 or a moderate ratio.\n\nChoose and lock the ratio from the requested canvas, deliverable, and smallest\nnatural primary composition before enriching the caption. Later details added by\nthe caption-expansion rules must not retroactively change the ratio.\n\nUniversal hard rules:\n1. Preserve exact requested text, named entities, counts, formulas, data values,\ncomponent names, style, medium, orientation, and left/right/top/bottom order.\n2. Required content must be visible, not implied. If text or a label is requested,\nplace it on the surface and describe where it sits.\n3. For active bodies, solve scene geometry before pose detail. Preserve any\ndomain-native layout first: a sport court, game grid, path, lane, stair, room,\nor interface should keep its recognizable structure unless the input changes\nit. Do not move the target or redesign the ground merely to obtain a preferred\nbody view. Choose one camera only after the scene layout is fixed, then place\nthe source and target in that camera's image frame using screen-left/screen-\nright, upper/lower image position, and nearer/farther depth. Silently trace the\nsource-to-target vector in this fixed view. Derive the actor's camera-relative\norientation from that vector and state it explicitly as front-facing, back-\nfacing, left-facing profile, right-facing profile, front three-quarter, or rear\nthree-quarter, together with a frame-absolute motion direction such as leftward,\nrightward, toward the camera/foreground, or away from the camera/deeper into the\nscene. Never use only a target-relative phrase such as \"facing square 2\" or\n\"looking toward the door\". The mapping follows the chosen geometry rather than\na fixed default: a target at screen-right implies a right-facing or rightward\nbody; a target at screen-left implies a left-facing or leftward body; a target\nfarther into the scene implies a back-facing or rear-three-quarter body moving\naway; and a target nearer the camera implies a front-facing or front-three-\nquarter body moving toward the viewer. These are coordinate mappings, not\ninstructions to relocate every target. Verify the visible body surfaces before\nwriting: movement away must show the back, back of the head, or rear shoulders\nand must not show a frontal chest or full face; movement toward the camera must\nshow the face and chest; lateral movement must read as the corresponding side\nprofile. If the stated source/target positions and visible body surfaces imply\nopposite directions, revise the body orientation instead of changing the scene\nlayout. Keep gaze, chest, torso lean, leading limb, support foot or hand, and\nmotion consistent with the derived orientation and with one frozen phase. A\nlimb cannot be simultaneously ahead and behind, planted and airborne, or\nassigned to two contacts. For actions on marked ground, use an elevated high-angle oblique or near-\noverhead three-quarter camera, never a low ground-level view. Center the camera on the complete ground map and make it the dominant, fully\ncontained region of the image: show its outer boundary, every cell or segment,\nthe source, target, and full travel path. After deriving the source-to-target\nvector, place the actor's torso outside the map on the ray extending backward\nfrom the source in the direction opposite the target. Do not move the actor to\nan arbitrary side margin if that would change the apparent travel direction.\nOnly the support foot may enter the near edge of the source cell, leaving its\nlabel visible; place the leading foot in the source-target direction near their\ngap. Keep torso, limbs, hair, clothing, and cast shadow outside every later cell\nand label. Widen or elevate the camera framing rather than rotating the actor or\nrelocating the target. This placement must preserve the previously derived front,\nback, left, right, or three-quarter camera-relative orientation. Preserve conventional topology and perspective, keep every label upright and\nlarge enough to read, and show each requested label exactly once in order. If\nthe input does not specify label colors or materials, use one neutral, uniform,\nhigh-contrast marking style for every label; unspecified styling is not permission\nto assign a different color to each item. Keep the actor's hands empty unless the\ninput names a held object. Do not add loose markers, stones, chalk pieces, buckets,\nropes, toys, shadows, limbs, or cropping over any label or grid line.\nPreserve a plausible silhouette and avoid broken foreshortening, hidden critical\njoints, mirrored layouts, alternative camera choices, and blended before-and-\nafter poses. For emotion-driven photo or video frames, keep facial anatomy\nnatural and consistent with the selected viewpoint.\n4. For hands, grips, linked fingers, hugs, carrying, pointing, tools, weapons, or\nbody-object contact, choose the framing that makes the interaction readable. Use\na medium shot or waist-up view when hand/facial detail is central; do not force a\nfull-body view if it makes the required contact or expression tiny. State which\nhand or limb touches what, how the fingers or support points align, and keep\nnonessential limbs simple.\n5. Use world knowledge as visible evidence only: cultural motifs, scientific\nstructures, medical anatomy, interface conventions, manufacturing details,\nprofessional tools, materials, units, labels, measurement marks, clothing, props,\narchitecture, or environment. Do not add lore that cannot be seen.\n6. Respect requested subject count. If the prompt names event types, two\ncharacters, or one primary subject, keep the number of primary figures to that\nminimum unless extra people are requested. Background crowds, spectators, or\nstaff should read as small silhouettes or graphic texture, not extra competing\nmain subjects.\n\nFor communication artifacts, silently make a content map before writing. Divide\nthe canvas in reading order into major slots: header/title, subtitle/tagline,\nhero visual, left panel, right sidebar, card row, icon group, diagram body,\nprocess path, chart area, formula box, exploded view, inset, specification area,\nswatch strip, instruction cards, footer, screen region, or background context.\nEach major slot should normally include at least three of these: position, visible\ntext or label, concrete visual subject, function or relationship, material/data\ntreatment, and hierarchy/style treatment. Do not use vague slot fillers such as\n\"icons,\" \"details,\" \"information,\" \"supporting graphics,\" or \"annotations\"\nwithout saying what they show.\n\nMake the artifact feel complete. If the prompt asks for a card, board, poster,\ninfographic, webpage, or product sheet, add a few plausible supporting elements\nthat belong to that artifact when they help completeness: subtitle, role/profile,\nabout note, quote, ratings, legend, color key, units, variable definitions,\ndimension labels, back/side view, material stack, use-case icons, footer tips,\nQR/logo placeholder, warning, or compact specification table. Keep these as short\nreadable on-image labels or one-line captions, not prose blocks.\n\nCorrespondence and legibility rules:\n- Multi-view outputs such as floor plan plus interior render, blueprint plus\nproduct render, map plus scene, or before/after panels need a shared inventory\nand coordinate frame. Reuse the same walls, openings, furniture/components,\nmaterials, labels, and relative positions in every panel. Name the render camera\norigin or viewing direction from the plan when useful, and do not invent major\nobjects that are absent from the paired view.\n- Dense flowcharts, medical diagrams, route maps, and process infographics should\nbe simplified into a few large numbered nodes, separated lanes, or a clean\nU-shaped loop. Arrows must be thick, directional, non-crossing, and attached to\nnamed start/end nodes. Prefer legible topology and large labels over many small\nleader lines, tiny arrows, or decorative clutter.\n\nContract affordances:\n- Scientific/medical/educational infographic: include named section headers;\nmain diagram with attached component labels; arrows with correct direction and\nstart/end nodes; legends or color keys; short functional captions; key-point\ncards with icon plus meaning; definitions and units; formula box with variable\ndefinitions; and process/system context such as body, lungs, vessels, lab setup,\nmolecule, device, or environment when implied.\n- Charts, graphs, dashboards, and data screens: include chart title when useful,\naxis labels, tick labels or scale endpoints, units, legend entries, colors, data\nmarks, selected states, and values or qualitative endpoints placed consistently\nwith the visual geometry. Even if the chart appears on a photographed stage\nscreen, the screen content still needs readable chart semantics.\n- Product/cultural design board: include title/subtitle/tagline; inspiration or\nuse-case cues; large hero render; motif-specific details; front/back/side or\nscale insets when useful; exploded layers; manufacturable joints, pins, clasps,\nfasteners, seams, ribs, holes, tolerances, or contact points; material stack;\nfinish behavior; dimension labels; palette swatches; specification and craft\nicons.\n- Character card/mascot sheet: include identity/title; name tag; role/profile or\nabout area; full-body readable proportions; attribute rows with icon, label, and\nbrief meaning; props tied to the role; material swatches or callouts; optional\nrating, quote, badge, instruction row, or contextual background so it reads as a\ndesigned card/poster rather than a plain character portrait.\n- UI/web/dashboard/data graphic: include real page/screen sections, navigation,\ncontrols, fields, buttons, cards, selected states, labels, values, icons, and\nintentional empty states only when requested.\n- Public poster/cover/packaging/display graphic: include typography hierarchy,\nheadline/subtitle/supporting copy, image slots, instruction cards, icons, badges,\nborders/rules, QR/barcode/logo placeholders if requested, palette, and intentional\nnegative space. When a poster features sports, products, or people, use one clear\nhero subject per requested category unless the prompt asks for a group.\n\nStyle comes after content. Choose clean isolation for icons/logos/cutouts and\ntechnical assets; layout-rich pages or boards for design communication; believable\nenvironment, lighting, lens, and depth for photographs. Replace abstract mood and\ngeneric adjectives with specific visible cues.\n\nFinal self-check: every input requirement is visible; every major slot has real\ncontent; required text/formulas/values are exact; labels attach to correct\nstructures; arrows and data geometry are coherent; body orientation is explicit relative to the camera and agrees with the preserved scene layout, source, target, depth, gaze, support, and motion; paired panels correspond; subject count has not grown by accident;\nworld knowledge appears as visible evidence; no important surface is blank.\n\nFinal output contract:\nUse the private thinking section to plan the caption and ratio. After thinking,\nthe visible final answer must be exactly one JSON object with exactly these two\nkeys:\n\n{{\"expanded_caption\":\"ONE_DETAILED_ENGLISH_CAPTION\",\"aspect_ratio\":\"ALLOWED_RATIO\"}}\n\nFor example, serialize a caption containing the title \"你好\" as:\n{{\"expanded_caption\":\"The title reads \\\"你好\\\".\",\"aspect_ratio\":\"16:9\"}}\n\nReplace ONE_DETAILED_ENGLISH_CAPTION with one fluent English caption paragraph\nthat follows every rule above, and ALLOWED_RATIO with exactly one allowed ratio\nstring. Keep expanded_caption as one valid JSON string. Every ASCII double quotation mark used as caption content inside expanded_caption must be serialized as the JSON escape sequence \\\". Escape all other characters as valid JSON requires. Do not add or omit keys. Do not put thinking or think tags inside\nthe object. Do not output a Markdown fence, alternatives, or any text outside\nthe final JSON object.\n\ninput_text: {input text}\noutput_text:\n"
266
+ ],
267
+ "title": "Upsampler task prompt"
268
+ },
269
+ {
270
+ "properties": {
271
+ "Node name for S&R": "StringReplace"
272
+ },
273
+ "id": 15,
274
+ "type": "StringReplace",
275
+ "pos": [
276
+ 470,
277
+ 900
278
+ ],
279
+ "size": [
280
+ 300,
281
+ 200
282
+ ],
283
+ "flags": {},
284
+ "order": 8,
285
+ "mode": 0,
286
+ "inputs": [
287
+ {
288
+ "name": "string",
289
+ "type": "STRING",
290
+ "link": 1,
291
+ "widget": {
292
+ "name": "string"
293
+ }
294
+ },
295
+ {
296
+ "name": "find",
297
+ "type": "STRING",
298
+ "link": null,
299
+ "widget": {
300
+ "name": "find"
301
+ }
302
+ },
303
+ {
304
+ "name": "replace",
305
+ "type": "STRING",
306
+ "link": 2,
307
+ "widget": {
308
+ "name": "replace"
309
+ }
310
+ }
311
+ ],
312
+ "outputs": [
313
+ {
314
+ "name": "STRING",
315
+ "type": "STRING",
316
+ "links": [
317
+ 5
318
+ ]
319
+ }
320
+ ],
321
+ "widgets_values": [
322
+ "",
323
+ "{input text}",
324
+ ""
325
+ ]
326
+ },
327
+ {
328
+ "properties": {
329
+ "Node name for S&R": "TextGenerateV2"
330
+ },
331
+ "id": 16,
332
+ "type": "TextGenerateV2",
333
+ "pos": [
334
+ 470,
335
+ 1160
336
+ ],
337
+ "size": [
338
+ 300,
339
+ 460
340
+ ],
341
+ "flags": {},
342
+ "order": 9,
343
+ "mode": 0,
344
+ "inputs": [
345
+ {
346
+ "name": "clip",
347
+ "type": "CLIP",
348
+ "link": 3
349
+ },
350
+ {
351
+ "name": "system_prompt",
352
+ "type": "STRING",
353
+ "link": 4
354
+ },
355
+ {
356
+ "name": "prompt",
357
+ "type": "STRING",
358
+ "link": 5,
359
+ "widget": {
360
+ "name": "prompt"
361
+ }
362
+ }
363
+ ],
364
+ "outputs": [
365
+ {
366
+ "name": "generated_text",
367
+ "type": "STRING",
368
+ "links": [
369
+ 6
370
+ ]
371
+ },
372
+ {
373
+ "name": "thinking",
374
+ "type": "STRING",
375
+ "links": []
376
+ }
377
+ ],
378
+ "widgets_values": [
379
+ "",
380
+ 8192,
381
+ "on",
382
+ 1,
383
+ 20,
384
+ 0.95,
385
+ 0,
386
+ 1,
387
+ 0,
388
+ 0,
389
+ true,
390
+ true,
391
+ "auto"
392
+ ]
393
+ },
394
+ {
395
+ "properties": {
396
+ "Node name for S&R": "JsonExtractString"
397
+ },
398
+ "id": 17,
399
+ "type": "JsonExtractString",
400
+ "pos": [
401
+ 0,
402
+ 1520
403
+ ],
404
+ "size": [
405
+ 420,
406
+ 110
407
+ ],
408
+ "flags": {},
409
+ "order": 10,
410
+ "mode": 0,
411
+ "inputs": [
412
+ {
413
+ "name": "json_string",
414
+ "type": "STRING",
415
+ "link": 6,
416
+ "widget": {
417
+ "name": "json_string"
418
+ }
419
+ },
420
+ {
421
+ "name": "key",
422
+ "type": "STRING",
423
+ "link": null,
424
+ "widget": {
425
+ "name": "key"
426
+ }
427
+ }
428
+ ],
429
+ "outputs": [
430
+ {
431
+ "name": "STRING",
432
+ "type": "STRING",
433
+ "links": [
434
+ 8
435
+ ]
436
+ }
437
+ ],
438
+ "widgets_values": [
439
+ "",
440
+ "expanded_caption"
441
+ ]
442
+ },
443
+ {
444
+ "properties": {
445
+ "Node name for S&R": "ComfySwitchNode"
446
+ },
447
+ "id": 18,
448
+ "type": "ComfySwitchNode",
449
+ "pos": [
450
+ 470,
451
+ 520
452
+ ],
453
+ "size": [
454
+ 300,
455
+ 90
456
+ ],
457
+ "flags": {},
458
+ "order": 11,
459
+ "mode": 0,
460
+ "inputs": [
461
+ {
462
+ "name": "on_false",
463
+ "type": "STRING",
464
+ "link": 7
465
+ },
466
+ {
467
+ "name": "on_true",
468
+ "type": "STRING",
469
+ "link": 8
470
+ },
471
+ {
472
+ "name": "switch",
473
+ "type": "BOOLEAN",
474
+ "link": 9,
475
+ "widget": {
476
+ "name": "switch"
477
+ }
478
+ }
479
+ ],
480
+ "outputs": [
481
+ {
482
+ "name": "output",
483
+ "type": "STRING",
484
+ "links": [
485
+ 10,
486
+ 27
487
+ ]
488
+ }
489
+ ],
490
+ "widgets_values": [
491
+ false
492
+ ],
493
+ "title": "Prompt: as written / enhanced"
494
+ },
495
+ {
496
+ "properties": {
497
+ "Node name for S&R": "PreviewAny"
498
+ },
499
+ "id": 19,
500
+ "type": "PreviewAny",
501
+ "pos": [
502
+ 470,
503
+ 660
504
+ ],
505
+ "size": [
506
+ 300,
507
+ 190
508
+ ],
509
+ "flags": {},
510
+ "order": 12,
511
+ "mode": 0,
512
+ "inputs": [
513
+ {
514
+ "name": "source",
515
+ "type": "*",
516
+ "link": 10
517
+ }
518
+ ],
519
+ "outputs": [
520
+ {
521
+ "name": "STRING",
522
+ "type": "STRING",
523
+ "links": []
524
+ }
525
+ ],
526
+ "widgets_values": [],
527
+ "title": "Prompt sent to the model"
528
+ },
529
+ {
530
+ "properties": {
531
+ "Node name for S&R": "PrimitiveBoolean"
532
+ },
533
+ "id": 20,
534
+ "type": "PrimitiveBoolean",
535
+ "pos": [
536
+ 840,
537
+ 0
538
+ ],
539
+ "size": [
540
+ 300,
541
+ 60
542
+ ],
543
+ "flags": {},
544
+ "order": 13,
545
+ "mode": 0,
546
+ "inputs": [],
547
+ "outputs": [
548
+ {
549
+ "name": "BOOLEAN",
550
+ "type": "BOOLEAN",
551
+ "links": [
552
+ 13
553
+ ]
554
+ }
555
+ ],
556
+ "widgets_values": [
557
+ true
558
+ ],
559
+ "title": "16 MP (off = 4 MP draft)"
560
+ },
561
+ {
562
+ "properties": {
563
+ "Node name for S&R": "PrimitiveFloat"
564
+ },
565
+ "id": 21,
566
+ "type": "PrimitiveFloat",
567
+ "pos": [
568
+ 840,
569
+ 110
570
+ ],
571
+ "size": [
572
+ 300,
573
+ 60
574
+ ],
575
+ "flags": {},
576
+ "order": 14,
577
+ "mode": 0,
578
+ "inputs": [],
579
+ "outputs": [
580
+ {
581
+ "name": "FLOAT",
582
+ "type": "FLOAT",
583
+ "links": [
584
+ 11
585
+ ]
586
+ }
587
+ ],
588
+ "widgets_values": [
589
+ 4.0
590
+ ],
591
+ "title": "Draft megapixels"
592
+ },
593
+ {
594
+ "properties": {
595
+ "Node name for S&R": "PrimitiveFloat"
596
+ },
597
+ "id": 22,
598
+ "type": "PrimitiveFloat",
599
+ "pos": [
600
+ 840,
601
+ 220
602
+ ],
603
+ "size": [
604
+ 300,
605
+ 60
606
+ ],
607
+ "flags": {},
608
+ "order": 15,
609
+ "mode": 0,
610
+ "inputs": [],
611
+ "outputs": [
612
+ {
613
+ "name": "FLOAT",
614
+ "type": "FLOAT",
615
+ "links": [
616
+ 12
617
+ ]
618
+ }
619
+ ],
620
+ "widgets_values": [
621
+ 16.0
622
+ ],
623
+ "title": "Full megapixels (16 GB VRAM: 16)"
624
+ },
625
+ {
626
+ "properties": {
627
+ "Node name for S&R": "ComfySwitchNode"
628
+ },
629
+ "id": 23,
630
+ "type": "ComfySwitchNode",
631
+ "pos": [
632
+ 840,
633
+ 330
634
+ ],
635
+ "size": [
636
+ 300,
637
+ 90
638
+ ],
639
+ "flags": {},
640
+ "order": 16,
641
+ "mode": 0,
642
+ "inputs": [
643
+ {
644
+ "name": "on_false",
645
+ "type": "FLOAT",
646
+ "link": 11
647
+ },
648
+ {
649
+ "name": "on_true",
650
+ "type": "FLOAT",
651
+ "link": 12
652
+ },
653
+ {
654
+ "name": "switch",
655
+ "type": "BOOLEAN",
656
+ "link": 13,
657
+ "widget": {
658
+ "name": "switch"
659
+ }
660
+ }
661
+ ],
662
+ "outputs": [
663
+ {
664
+ "name": "output",
665
+ "type": "FLOAT",
666
+ "links": [
667
+ 14
668
+ ]
669
+ }
670
+ ],
671
+ "widgets_values": [
672
+ false
673
+ ],
674
+ "title": "Megapixels: draft / 16 MP"
675
+ },
676
+ {
677
+ "properties": {
678
+ "Node name for S&R": "ResolutionSelector"
679
+ },
680
+ "id": 24,
681
+ "type": "ResolutionSelector",
682
+ "pos": [
683
+ 1190,
684
+ 0
685
+ ],
686
+ "size": [
687
+ 300,
688
+ 240
689
+ ],
690
+ "flags": {},
691
+ "order": 17,
692
+ "mode": 0,
693
+ "inputs": [
694
+ {
695
+ "name": "aspect_ratio",
696
+ "type": "COMBO",
697
+ "link": null,
698
+ "widget": {
699
+ "name": "aspect_ratio"
700
+ }
701
+ },
702
+ {
703
+ "name": "megapixels",
704
+ "type": "FLOAT",
705
+ "link": 14,
706
+ "widget": {
707
+ "name": "megapixels"
708
+ }
709
+ },
710
+ {
711
+ "name": "multiple",
712
+ "type": "INT",
713
+ "link": null,
714
+ "widget": {
715
+ "name": "multiple"
716
+ }
717
+ }
718
+ ],
719
+ "outputs": [
720
+ {
721
+ "name": "width",
722
+ "type": "INT",
723
+ "links": [
724
+ 28
725
+ ]
726
+ },
727
+ {
728
+ "name": "height",
729
+ "type": "INT",
730
+ "links": [
731
+ 29
732
+ ]
733
+ }
734
+ ],
735
+ "widgets_values": [
736
+ "16:9 (Widescreen)",
737
+ 16.0,
738
+ 32
739
+ ],
740
+ "title": "ASPECT RATIO"
741
+ },
742
+ {
743
+ "properties": {
744
+ "Node name for S&R": "MiniMaxH3ImageConditioning"
745
+ },
746
+ "id": 30,
747
+ "type": "MiniMaxH3ImageConditioning",
748
+ "pos": [
749
+ 1560,
750
+ 0
751
+ ],
752
+ "size": [
753
+ 330,
754
+ 400
755
+ ],
756
+ "flags": {},
757
+ "order": 18,
758
+ "mode": 0,
759
+ "inputs": [
760
+ {
761
+ "name": "clip",
762
+ "type": "CLIP",
763
+ "link": 15
764
+ },
765
+ {
766
+ "name": "vae",
767
+ "type": "VAE",
768
+ "link": 16
769
+ },
770
+ {
771
+ "name": "images.image_1",
772
+ "type": "IMAGE",
773
+ "link": 17
774
+ },
775
+ {
776
+ "name": "images.image_2",
777
+ "type": "IMAGE",
778
+ "link": 18
779
+ },
780
+ {
781
+ "name": "images.image_3",
782
+ "type": "IMAGE",
783
+ "link": 19
784
+ },
785
+ {
786
+ "name": "images.image_4",
787
+ "type": "IMAGE",
788
+ "link": 20
789
+ },
790
+ {
791
+ "name": "images.image_5",
792
+ "type": "IMAGE",
793
+ "link": 21
794
+ },
795
+ {
796
+ "name": "images.image_6",
797
+ "type": "IMAGE",
798
+ "link": 22
799
+ },
800
+ {
801
+ "name": "images.image_7",
802
+ "type": "IMAGE",
803
+ "link": 23
804
+ },
805
+ {
806
+ "name": "images.image_8",
807
+ "type": "IMAGE",
808
+ "link": 24
809
+ },
810
+ {
811
+ "name": "images.image_9",
812
+ "type": "IMAGE",
813
+ "link": 25
814
+ },
815
+ {
816
+ "name": "images.image_10",
817
+ "type": "IMAGE",
818
+ "link": 26
819
+ },
820
+ {
821
+ "name": "prompt",
822
+ "type": "STRING",
823
+ "link": 27,
824
+ "widget": {
825
+ "name": "prompt"
826
+ }
827
+ },
828
+ {
829
+ "name": "width",
830
+ "type": "INT",
831
+ "link": 28,
832
+ "widget": {
833
+ "name": "width"
834
+ }
835
+ },
836
+ {
837
+ "name": "height",
838
+ "type": "INT",
839
+ "link": 29,
840
+ "widget": {
841
+ "name": "height"
842
+ }
843
+ }
844
+ ],
845
+ "outputs": [
846
+ {
847
+ "name": "positive",
848
+ "type": "CONDITIONING",
849
+ "links": [
850
+ 30,
851
+ 34
852
+ ]
853
+ },
854
+ {
855
+ "name": "LATENT",
856
+ "type": "LATENT",
857
+ "links": [
858
+ 41
859
+ ]
860
+ }
861
+ ],
862
+ "widgets_values": [
863
+ "",
864
+ 0,
865
+ 0
866
+ ]
867
+ },
868
+ {
869
+ "properties": {
870
+ "Node name for S&R": "ConditioningZeroOut"
871
+ },
872
+ "id": 31,
873
+ "type": "ConditioningZeroOut",
874
+ "pos": [
875
+ 1940,
876
+ 0
877
+ ],
878
+ "size": [
879
+ 300,
880
+ 50
881
+ ],
882
+ "flags": {},
883
+ "order": 19,
884
+ "mode": 0,
885
+ "inputs": [
886
+ {
887
+ "name": "conditioning",
888
+ "type": "CONDITIONING",
889
+ "link": 30
890
+ }
891
+ ],
892
+ "outputs": [
893
+ {
894
+ "name": "CONDITIONING",
895
+ "type": "CONDITIONING",
896
+ "links": [
897
+ 35
898
+ ]
899
+ }
900
+ ]
901
+ },
902
+ {
903
+ "properties": {
904
+ "Node name for S&R": "VAEDecode"
905
+ },
906
+ "id": 33,
907
+ "type": "VAEDecode",
908
+ "pos": [
909
+ 2290,
910
+ 0
911
+ ],
912
+ "size": [
913
+ 220,
914
+ 50
915
+ ],
916
+ "flags": {},
917
+ "order": 20,
918
+ "mode": 0,
919
+ "inputs": [
920
+ {
921
+ "name": "samples",
922
+ "type": "LATENT",
923
+ "link": 42
924
+ },
925
+ {
926
+ "name": "vae",
927
+ "type": "VAE",
928
+ "link": 31
929
+ }
930
+ ],
931
+ "outputs": [
932
+ {
933
+ "name": "IMAGE",
934
+ "type": "IMAGE",
935
+ "links": [
936
+ 32
937
+ ]
938
+ }
939
+ ]
940
+ },
941
+ {
942
+ "properties": {
943
+ "Node name for S&R": "SaveImage"
944
+ },
945
+ "id": 34,
946
+ "type": "SaveImage",
947
+ "pos": [
948
+ 2290,
949
+ 100
950
+ ],
951
+ "size": [
952
+ 560,
953
+ 520
954
+ ],
955
+ "flags": {},
956
+ "order": 21,
957
+ "mode": 0,
958
+ "inputs": [
959
+ {
960
+ "name": "images",
961
+ "type": "IMAGE",
962
+ "link": 32
963
+ }
964
+ ],
965
+ "outputs": [],
966
+ "widgets_values": [
967
+ "h3_image/minimax_h3_image_16gb"
968
+ ]
969
+ },
970
+ {
971
+ "properties": {
972
+ "Node name for S&R": "RandomNoise"
973
+ },
974
+ "id": 35,
975
+ "type": "RandomNoise",
976
+ "pos": [
977
+ 1940,
978
+ 90
979
+ ],
980
+ "size": [
981
+ 300,
982
+ 82
983
+ ],
984
+ "flags": {},
985
+ "order": 22,
986
+ "mode": 0,
987
+ "inputs": [
988
+ {
989
+ "name": "noise_seed",
990
+ "type": "INT",
991
+ "link": null,
992
+ "widget": {
993
+ "name": "noise_seed"
994
+ }
995
+ }
996
+ ],
997
+ "outputs": [
998
+ {
999
+ "name": "NOISE",
1000
+ "type": "NOISE",
1001
+ "links": [
1002
+ 37
1003
+ ]
1004
+ }
1005
+ ],
1006
+ "widgets_values": [
1007
+ 42,
1008
+ "randomize"
1009
+ ],
1010
+ "title": "Seed"
1011
+ },
1012
+ {
1013
+ "properties": {
1014
+ "Node name for S&R": "CFGGuider"
1015
+ },
1016
+ "id": 36,
1017
+ "type": "CFGGuider",
1018
+ "pos": [
1019
+ 1940,
1020
+ 200
1021
+ ],
1022
+ "size": [
1023
+ 300,
1024
+ 100
1025
+ ],
1026
+ "flags": {},
1027
+ "order": 23,
1028
+ "mode": 0,
1029
+ "inputs": [
1030
+ {
1031
+ "name": "model",
1032
+ "type": "MODEL",
1033
+ "link": 33
1034
+ },
1035
+ {
1036
+ "name": "positive",
1037
+ "type": "CONDITIONING",
1038
+ "link": 34
1039
+ },
1040
+ {
1041
+ "name": "negative",
1042
+ "type": "CONDITIONING",
1043
+ "link": 35
1044
+ },
1045
+ {
1046
+ "name": "cfg",
1047
+ "type": "FLOAT",
1048
+ "link": null,
1049
+ "widget": {
1050
+ "name": "cfg"
1051
+ }
1052
+ }
1053
+ ],
1054
+ "outputs": [
1055
+ {
1056
+ "name": "GUIDER",
1057
+ "type": "GUIDER",
1058
+ "links": [
1059
+ 38
1060
+ ]
1061
+ }
1062
+ ],
1063
+ "widgets_values": [
1064
+ 1.0
1065
+ ]
1066
+ },
1067
+ {
1068
+ "properties": {
1069
+ "Node name for S&R": "BasicScheduler"
1070
+ },
1071
+ "id": 37,
1072
+ "type": "BasicScheduler",
1073
+ "pos": [
1074
+ 1940,
1075
+ 330
1076
+ ],
1077
+ "size": [
1078
+ 300,
1079
+ 110
1080
+ ],
1081
+ "flags": {},
1082
+ "order": 24,
1083
+ "mode": 0,
1084
+ "inputs": [
1085
+ {
1086
+ "name": "model",
1087
+ "type": "MODEL",
1088
+ "link": 36
1089
+ },
1090
+ {
1091
+ "name": "scheduler",
1092
+ "type": "COMBO",
1093
+ "link": null,
1094
+ "widget": {
1095
+ "name": "scheduler"
1096
+ }
1097
+ },
1098
+ {
1099
+ "name": "steps",
1100
+ "type": "INT",
1101
+ "link": null,
1102
+ "widget": {
1103
+ "name": "steps"
1104
+ }
1105
+ },
1106
+ {
1107
+ "name": "denoise",
1108
+ "type": "FLOAT",
1109
+ "link": null,
1110
+ "widget": {
1111
+ "name": "denoise"
1112
+ }
1113
+ }
1114
+ ],
1115
+ "outputs": [
1116
+ {
1117
+ "name": "SIGMAS",
1118
+ "type": "SIGMAS",
1119
+ "links": [
1120
+ 40
1121
+ ]
1122
+ }
1123
+ ],
1124
+ "widgets_values": [
1125
+ "simple",
1126
+ 6,
1127
+ 1.0
1128
+ ]
1129
+ },
1130
+ {
1131
+ "properties": {
1132
+ "Node name for S&R": "SamplerLCM"
1133
+ },
1134
+ "id": 38,
1135
+ "type": "SamplerLCM",
1136
+ "pos": [
1137
+ 1940,
1138
+ 470
1139
+ ],
1140
+ "size": [
1141
+ 300,
1142
+ 110
1143
+ ],
1144
+ "flags": {},
1145
+ "order": 25,
1146
+ "mode": 0,
1147
+ "inputs": [
1148
+ {
1149
+ "name": "s_noise",
1150
+ "type": "FLOAT",
1151
+ "link": null,
1152
+ "widget": {
1153
+ "name": "s_noise"
1154
+ }
1155
+ },
1156
+ {
1157
+ "name": "s_noise_end",
1158
+ "type": "FLOAT",
1159
+ "link": null,
1160
+ "widget": {
1161
+ "name": "s_noise_end"
1162
+ }
1163
+ },
1164
+ {
1165
+ "name": "noise_clip_std",
1166
+ "type": "FLOAT",
1167
+ "link": null,
1168
+ "widget": {
1169
+ "name": "noise_clip_std"
1170
+ }
1171
+ }
1172
+ ],
1173
+ "outputs": [
1174
+ {
1175
+ "name": "SAMPLER",
1176
+ "type": "SAMPLER",
1177
+ "links": [
1178
+ 39
1179
+ ]
1180
+ }
1181
+ ],
1182
+ "widgets_values": [
1183
+ 1.1,
1184
+ 0.9,
1185
+ 0.0
1186
+ ],
1187
+ "title": "SamplerLCM (real skin)"
1188
+ },
1189
+ {
1190
+ "properties": {
1191
+ "Node name for S&R": "SamplerCustomAdvanced"
1192
+ },
1193
+ "id": 39,
1194
+ "type": "SamplerCustomAdvanced",
1195
+ "pos": [
1196
+ 1560,
1197
+ 440
1198
+ ],
1199
+ "size": [
1200
+ 330,
1201
+ 130
1202
+ ],
1203
+ "flags": {},
1204
+ "order": 26,
1205
+ "mode": 0,
1206
+ "inputs": [
1207
+ {
1208
+ "name": "noise",
1209
+ "type": "NOISE",
1210
+ "link": 37
1211
+ },
1212
+ {
1213
+ "name": "guider",
1214
+ "type": "GUIDER",
1215
+ "link": 38
1216
+ },
1217
+ {
1218
+ "name": "sampler",
1219
+ "type": "SAMPLER",
1220
+ "link": 39
1221
+ },
1222
+ {
1223
+ "name": "sigmas",
1224
+ "type": "SIGMAS",
1225
+ "link": 40
1226
+ },
1227
+ {
1228
+ "name": "latent_image",
1229
+ "type": "LATENT",
1230
+ "link": 41
1231
+ }
1232
+ ],
1233
+ "outputs": [
1234
+ {
1235
+ "name": "output",
1236
+ "type": "LATENT",
1237
+ "links": [
1238
+ 42
1239
+ ]
1240
+ },
1241
+ {
1242
+ "name": "denoised_output",
1243
+ "type": "LATENT",
1244
+ "links": []
1245
+ }
1246
+ ]
1247
+ },
1248
+ {
1249
+ "properties": {
1250
+ "Node name for S&R": "LoadImage"
1251
+ },
1252
+ "id": 41,
1253
+ "type": "LoadImage",
1254
+ "pos": [
1255
+ 840,
1256
+ 790
1257
+ ],
1258
+ "size": [
1259
+ 270,
1260
+ 360
1261
+ ],
1262
+ "flags": {},
1263
+ "order": 27,
1264
+ "mode": 4,
1265
+ "inputs": [],
1266
+ "outputs": [
1267
+ {
1268
+ "name": "IMAGE",
1269
+ "type": "IMAGE",
1270
+ "links": [
1271
+ 17
1272
+ ]
1273
+ },
1274
+ {
1275
+ "name": "MASK",
1276
+ "type": "MASK",
1277
+ "links": []
1278
+ }
1279
+ ],
1280
+ "widgets_values": [
1281
+ "01_Elena.png",
1282
+ "image"
1283
+ ],
1284
+ "title": "Picture 1 (the one to edit)"
1285
+ },
1286
+ {
1287
+ "properties": {
1288
+ "Node name for S&R": "LoadImage"
1289
+ },
1290
+ "id": 42,
1291
+ "type": "LoadImage",
1292
+ "pos": [
1293
+ 1130,
1294
+ 790
1295
+ ],
1296
+ "size": [
1297
+ 270,
1298
+ 360
1299
+ ],
1300
+ "flags": {},
1301
+ "order": 28,
1302
+ "mode": 4,
1303
+ "inputs": [],
1304
+ "outputs": [
1305
+ {
1306
+ "name": "IMAGE",
1307
+ "type": "IMAGE",
1308
+ "links": [
1309
+ 18
1310
+ ]
1311
+ },
1312
+ {
1313
+ "name": "MASK",
1314
+ "type": "MASK",
1315
+ "links": []
1316
+ }
1317
+ ],
1318
+ "widgets_values": [
1319
+ "01_Elena.png",
1320
+ "image"
1321
+ ],
1322
+ "title": "Picture 2"
1323
+ },
1324
+ {
1325
+ "properties": {
1326
+ "Node name for S&R": "LoadImage"
1327
+ },
1328
+ "id": 43,
1329
+ "type": "LoadImage",
1330
+ "pos": [
1331
+ 1420,
1332
+ 790
1333
+ ],
1334
+ "size": [
1335
+ 270,
1336
+ 360
1337
+ ],
1338
+ "flags": {},
1339
+ "order": 29,
1340
+ "mode": 4,
1341
+ "inputs": [],
1342
+ "outputs": [
1343
+ {
1344
+ "name": "IMAGE",
1345
+ "type": "IMAGE",
1346
+ "links": [
1347
+ 19
1348
+ ]
1349
+ },
1350
+ {
1351
+ "name": "MASK",
1352
+ "type": "MASK",
1353
+ "links": []
1354
+ }
1355
+ ],
1356
+ "widgets_values": [
1357
+ "01_Elena.png",
1358
+ "image"
1359
+ ],
1360
+ "title": "Picture 3"
1361
+ },
1362
+ {
1363
+ "properties": {
1364
+ "Node name for S&R": "LoadImage"
1365
+ },
1366
+ "id": 44,
1367
+ "type": "LoadImage",
1368
+ "pos": [
1369
+ 1710,
1370
+ 790
1371
+ ],
1372
+ "size": [
1373
+ 270,
1374
+ 360
1375
+ ],
1376
+ "flags": {},
1377
+ "order": 30,
1378
+ "mode": 4,
1379
+ "inputs": [],
1380
+ "outputs": [
1381
+ {
1382
+ "name": "IMAGE",
1383
+ "type": "IMAGE",
1384
+ "links": [
1385
+ 20
1386
+ ]
1387
+ },
1388
+ {
1389
+ "name": "MASK",
1390
+ "type": "MASK",
1391
+ "links": []
1392
+ }
1393
+ ],
1394
+ "widgets_values": [
1395
+ "01_Elena.png",
1396
+ "image"
1397
+ ],
1398
+ "title": "Picture 4"
1399
+ },
1400
+ {
1401
+ "properties": {
1402
+ "Node name for S&R": "LoadImage"
1403
+ },
1404
+ "id": 45,
1405
+ "type": "LoadImage",
1406
+ "pos": [
1407
+ 2000,
1408
+ 790
1409
+ ],
1410
+ "size": [
1411
+ 270,
1412
+ 360
1413
+ ],
1414
+ "flags": {},
1415
+ "order": 31,
1416
+ "mode": 4,
1417
+ "inputs": [],
1418
+ "outputs": [
1419
+ {
1420
+ "name": "IMAGE",
1421
+ "type": "IMAGE",
1422
+ "links": [
1423
+ 21
1424
+ ]
1425
+ },
1426
+ {
1427
+ "name": "MASK",
1428
+ "type": "MASK",
1429
+ "links": []
1430
+ }
1431
+ ],
1432
+ "widgets_values": [
1433
+ "01_Elena.png",
1434
+ "image"
1435
+ ],
1436
+ "title": "Picture 5"
1437
+ },
1438
+ {
1439
+ "properties": {
1440
+ "Node name for S&R": "LoadImage"
1441
+ },
1442
+ "id": 46,
1443
+ "type": "LoadImage",
1444
+ "pos": [
1445
+ 840,
1446
+ 1210
1447
+ ],
1448
+ "size": [
1449
+ 270,
1450
+ 360
1451
+ ],
1452
+ "flags": {},
1453
+ "order": 32,
1454
+ "mode": 4,
1455
+ "inputs": [],
1456
+ "outputs": [
1457
+ {
1458
+ "name": "IMAGE",
1459
+ "type": "IMAGE",
1460
+ "links": [
1461
+ 22
1462
+ ]
1463
+ },
1464
+ {
1465
+ "name": "MASK",
1466
+ "type": "MASK",
1467
+ "links": []
1468
+ }
1469
+ ],
1470
+ "widgets_values": [
1471
+ "01_Elena.png",
1472
+ "image"
1473
+ ],
1474
+ "title": "Picture 6"
1475
+ },
1476
+ {
1477
+ "properties": {
1478
+ "Node name for S&R": "LoadImage"
1479
+ },
1480
+ "id": 47,
1481
+ "type": "LoadImage",
1482
+ "pos": [
1483
+ 1130,
1484
+ 1210
1485
+ ],
1486
+ "size": [
1487
+ 270,
1488
+ 360
1489
+ ],
1490
+ "flags": {},
1491
+ "order": 33,
1492
+ "mode": 4,
1493
+ "inputs": [],
1494
+ "outputs": [
1495
+ {
1496
+ "name": "IMAGE",
1497
+ "type": "IMAGE",
1498
+ "links": [
1499
+ 23
1500
+ ]
1501
+ },
1502
+ {
1503
+ "name": "MASK",
1504
+ "type": "MASK",
1505
+ "links": []
1506
+ }
1507
+ ],
1508
+ "widgets_values": [
1509
+ "01_Elena.png",
1510
+ "image"
1511
+ ],
1512
+ "title": "Picture 7"
1513
+ },
1514
+ {
1515
+ "properties": {
1516
+ "Node name for S&R": "LoadImage"
1517
+ },
1518
+ "id": 48,
1519
+ "type": "LoadImage",
1520
+ "pos": [
1521
+ 1420,
1522
+ 1210
1523
+ ],
1524
+ "size": [
1525
+ 270,
1526
+ 360
1527
+ ],
1528
+ "flags": {},
1529
+ "order": 34,
1530
+ "mode": 4,
1531
+ "inputs": [],
1532
+ "outputs": [
1533
+ {
1534
+ "name": "IMAGE",
1535
+ "type": "IMAGE",
1536
+ "links": [
1537
+ 24
1538
+ ]
1539
+ },
1540
+ {
1541
+ "name": "MASK",
1542
+ "type": "MASK",
1543
+ "links": []
1544
+ }
1545
+ ],
1546
+ "widgets_values": [
1547
+ "01_Elena.png",
1548
+ "image"
1549
+ ],
1550
+ "title": "Picture 8"
1551
+ },
1552
+ {
1553
+ "properties": {
1554
+ "Node name for S&R": "LoadImage"
1555
+ },
1556
+ "id": 49,
1557
+ "type": "LoadImage",
1558
+ "pos": [
1559
+ 1710,
1560
+ 1210
1561
+ ],
1562
+ "size": [
1563
+ 270,
1564
+ 360
1565
+ ],
1566
+ "flags": {},
1567
+ "order": 35,
1568
+ "mode": 4,
1569
+ "inputs": [],
1570
+ "outputs": [
1571
+ {
1572
+ "name": "IMAGE",
1573
+ "type": "IMAGE",
1574
+ "links": [
1575
+ 25
1576
+ ]
1577
+ },
1578
+ {
1579
+ "name": "MASK",
1580
+ "type": "MASK",
1581
+ "links": []
1582
+ }
1583
+ ],
1584
+ "widgets_values": [
1585
+ "01_Elena.png",
1586
+ "image"
1587
+ ],
1588
+ "title": "Picture 9"
1589
+ },
1590
+ {
1591
+ "properties": {
1592
+ "Node name for S&R": "LoadImage"
1593
+ },
1594
+ "id": 50,
1595
+ "type": "LoadImage",
1596
+ "pos": [
1597
+ 2000,
1598
+ 1210
1599
+ ],
1600
+ "size": [
1601
+ 270,
1602
+ 360
1603
+ ],
1604
+ "flags": {},
1605
+ "order": 36,
1606
+ "mode": 4,
1607
+ "inputs": [],
1608
+ "outputs": [
1609
+ {
1610
+ "name": "IMAGE",
1611
+ "type": "IMAGE",
1612
+ "links": [
1613
+ 26
1614
+ ]
1615
+ },
1616
+ {
1617
+ "name": "MASK",
1618
+ "type": "MASK",
1619
+ "links": []
1620
+ }
1621
+ ],
1622
+ "widgets_values": [
1623
+ "01_Elena.png",
1624
+ "image"
1625
+ ],
1626
+ "title": "Picture 10"
1627
+ },
1628
+ {
1629
+ "properties": {
1630
+ "Node name for S&R": "MarkdownNote"
1631
+ },
1632
+ "id": 60,
1633
+ "type": "MarkdownNote",
1634
+ "pos": [
1635
+ -570,
1636
+ 0
1637
+ ],
1638
+ "size": [
1639
+ 520,
1640
+ 1100
1641
+ ],
1642
+ "flags": {},
1643
+ "order": 37,
1644
+ "mode": 0,
1645
+ "inputs": [],
1646
+ "outputs": [],
1647
+ "widgets_values": [
1648
+ "## MiniMax H3 Image - 16 GB VRAM version\n\nSame model and sampler as the 16 MP workflow, sized for a **16 GB** card: **16 MP** final, **4 MP** draft. The model was trained on images up to 16 MP; more megapixels give more skin and face detail, but need more VRAM.\n\n**1. Prompt** - write it **long**, like a real photo: who (age, hair, skin, outfit), where, what is around them, the light and its direction, and the camera, e.g. *\"Shot on the iPhone 16 Pro front camera, visible skin texture and pores, natural phone HDR, no retouching, ordinary Instagram photo, not cinematic, not studio.\"* Short prompts give a glossy doll face. *Enhance prompt* (Qwen3.5 upsampler) can write the long version for text-to-image; leave it off with reference pictures.\n\n**2. Size** - pick the ASPECT RATIO. *16 MP* on = final, off = 4 MP draft for testing prompts fast. People: use 3:4 (or 4:3). To try a bigger size, raise *Full megapixels* step by step (max 16).\n\n| Ratio | draft 4 MP | 16 MP |\n|---|---|---|\n| 1:1 | 2048x2048 | 4096x4096 |\n| 3:4 | 1760x2368 | 3552x4736 |\n| 4:3 | 2368x1760 | 4736x3552 |\n| 2:3 | 1664x2496 | 3360x5024 |\n| 3:2 | 2496x1664 | 5024x3360 |\n| 9:16 | 1536x2720 | 3072x5472 |\n| 16:9 | 2720x1536 | 5472x3072 |\n| 21:9 | 3136x1344 | 6272x2688 |\n\n**3. Reference pictures** - ten slots, all bypassed. Select a slot, press **Ctrl+B** to switch it on, and load your picture. Fill slots from 1 upward: the pictures that are on are numbered in slot order, so with slots 1, 2 and 3 on they are `<Picture 1>`, `<Picture 2>`, `<Picture 3>` in the prompt. Picture 1 is the one an edit instruction changes. Each picture adds time and memory: on 16 GB keep it to **10 pictures or fewer**. Keep the VAE connected to the conditioning node so the face and room do not drift.\n\n**4. Sampler** - SamplerCustomAdvanced with **SamplerLCM** (s_noise 1.1 -> s_noise_end 0.9, noise_clip_std 0), BasicScheduler **simple, 6 steps**, denoise 1.0, CFGGuider **cfg 1.0** (no negative prompt). Do not raise the steps: the model is distilled for 6.\n\n**Decode** - plain VAE Decode. It falls back to tiled decode by itself if VRAM runs out.\n"
1649
+ ],
1650
+ "title": "How to use (16 GB VRAM)"
1651
+ }
1652
+ ],
1653
+ "links": [
1654
+ [
1655
+ 1,
1656
+ 14,
1657
+ 0,
1658
+ 15,
1659
+ 0,
1660
+ "STRING"
1661
+ ],
1662
+ [
1663
+ 2,
1664
+ 10,
1665
+ 0,
1666
+ 15,
1667
+ 2,
1668
+ "STRING"
1669
+ ],
1670
+ [
1671
+ 3,
1672
+ 12,
1673
+ 0,
1674
+ 16,
1675
+ 0,
1676
+ "CLIP"
1677
+ ],
1678
+ [
1679
+ 4,
1680
+ 13,
1681
+ 0,
1682
+ 16,
1683
+ 1,
1684
+ "STRING"
1685
+ ],
1686
+ [
1687
+ 5,
1688
+ 15,
1689
+ 0,
1690
+ 16,
1691
+ 2,
1692
+ "STRING"
1693
+ ],
1694
+ [
1695
+ 6,
1696
+ 16,
1697
+ 0,
1698
+ 17,
1699
+ 0,
1700
+ "STRING"
1701
+ ],
1702
+ [
1703
+ 7,
1704
+ 10,
1705
+ 0,
1706
+ 18,
1707
+ 0,
1708
+ "STRING"
1709
+ ],
1710
+ [
1711
+ 8,
1712
+ 17,
1713
+ 0,
1714
+ 18,
1715
+ 1,
1716
+ "STRING"
1717
+ ],
1718
+ [
1719
+ 9,
1720
+ 11,
1721
+ 0,
1722
+ 18,
1723
+ 2,
1724
+ "BOOLEAN"
1725
+ ],
1726
+ [
1727
+ 10,
1728
+ 18,
1729
+ 0,
1730
+ 19,
1731
+ 0,
1732
+ "STRING"
1733
+ ],
1734
+ [
1735
+ 11,
1736
+ 21,
1737
+ 0,
1738
+ 23,
1739
+ 0,
1740
+ "FLOAT"
1741
+ ],
1742
+ [
1743
+ 12,
1744
+ 22,
1745
+ 0,
1746
+ 23,
1747
+ 1,
1748
+ "FLOAT"
1749
+ ],
1750
+ [
1751
+ 13,
1752
+ 20,
1753
+ 0,
1754
+ 23,
1755
+ 2,
1756
+ "BOOLEAN"
1757
+ ],
1758
+ [
1759
+ 14,
1760
+ 23,
1761
+ 0,
1762
+ 24,
1763
+ 1,
1764
+ "FLOAT"
1765
+ ],
1766
+ [
1767
+ 15,
1768
+ 2,
1769
+ 0,
1770
+ 30,
1771
+ 0,
1772
+ "CLIP"
1773
+ ],
1774
+ [
1775
+ 16,
1776
+ 3,
1777
+ 0,
1778
+ 30,
1779
+ 1,
1780
+ "VAE"
1781
+ ],
1782
+ [
1783
+ 17,
1784
+ 41,
1785
+ 0,
1786
+ 30,
1787
+ 2,
1788
+ "IMAGE"
1789
+ ],
1790
+ [
1791
+ 18,
1792
+ 42,
1793
+ 0,
1794
+ 30,
1795
+ 3,
1796
+ "IMAGE"
1797
+ ],
1798
+ [
1799
+ 19,
1800
+ 43,
1801
+ 0,
1802
+ 30,
1803
+ 4,
1804
+ "IMAGE"
1805
+ ],
1806
+ [
1807
+ 20,
1808
+ 44,
1809
+ 0,
1810
+ 30,
1811
+ 5,
1812
+ "IMAGE"
1813
+ ],
1814
+ [
1815
+ 21,
1816
+ 45,
1817
+ 0,
1818
+ 30,
1819
+ 6,
1820
+ "IMAGE"
1821
+ ],
1822
+ [
1823
+ 22,
1824
+ 46,
1825
+ 0,
1826
+ 30,
1827
+ 7,
1828
+ "IMAGE"
1829
+ ],
1830
+ [
1831
+ 23,
1832
+ 47,
1833
+ 0,
1834
+ 30,
1835
+ 8,
1836
+ "IMAGE"
1837
+ ],
1838
+ [
1839
+ 24,
1840
+ 48,
1841
+ 0,
1842
+ 30,
1843
+ 9,
1844
+ "IMAGE"
1845
+ ],
1846
+ [
1847
+ 25,
1848
+ 49,
1849
+ 0,
1850
+ 30,
1851
+ 10,
1852
+ "IMAGE"
1853
+ ],
1854
+ [
1855
+ 26,
1856
+ 50,
1857
+ 0,
1858
+ 30,
1859
+ 11,
1860
+ "IMAGE"
1861
+ ],
1862
+ [
1863
+ 27,
1864
+ 18,
1865
+ 0,
1866
+ 30,
1867
+ 12,
1868
+ "STRING"
1869
+ ],
1870
+ [
1871
+ 28,
1872
+ 24,
1873
+ 0,
1874
+ 30,
1875
+ 13,
1876
+ "INT"
1877
+ ],
1878
+ [
1879
+ 29,
1880
+ 24,
1881
+ 1,
1882
+ 30,
1883
+ 14,
1884
+ "INT"
1885
+ ],
1886
+ [
1887
+ 30,
1888
+ 30,
1889
+ 0,
1890
+ 31,
1891
+ 0,
1892
+ "CONDITIONING"
1893
+ ],
1894
+ [
1895
+ 31,
1896
+ 3,
1897
+ 0,
1898
+ 33,
1899
+ 1,
1900
+ "VAE"
1901
+ ],
1902
+ [
1903
+ 32,
1904
+ 33,
1905
+ 0,
1906
+ 34,
1907
+ 0,
1908
+ "IMAGE"
1909
+ ],
1910
+ [
1911
+ 33,
1912
+ 1,
1913
+ 0,
1914
+ 36,
1915
+ 0,
1916
+ "MODEL"
1917
+ ],
1918
+ [
1919
+ 34,
1920
+ 30,
1921
+ 0,
1922
+ 36,
1923
+ 1,
1924
+ "CONDITIONING"
1925
+ ],
1926
+ [
1927
+ 35,
1928
+ 31,
1929
+ 0,
1930
+ 36,
1931
+ 2,
1932
+ "CONDITIONING"
1933
+ ],
1934
+ [
1935
+ 36,
1936
+ 1,
1937
+ 0,
1938
+ 37,
1939
+ 0,
1940
+ "MODEL"
1941
+ ],
1942
+ [
1943
+ 37,
1944
+ 35,
1945
+ 0,
1946
+ 39,
1947
+ 0,
1948
+ "NOISE"
1949
+ ],
1950
+ [
1951
+ 38,
1952
+ 36,
1953
+ 0,
1954
+ 39,
1955
+ 1,
1956
+ "GUIDER"
1957
+ ],
1958
+ [
1959
+ 39,
1960
+ 38,
1961
+ 0,
1962
+ 39,
1963
+ 2,
1964
+ "SAMPLER"
1965
+ ],
1966
+ [
1967
+ 40,
1968
+ 37,
1969
+ 0,
1970
+ 39,
1971
+ 3,
1972
+ "SIGMAS"
1973
+ ],
1974
+ [
1975
+ 41,
1976
+ 30,
1977
+ 1,
1978
+ 39,
1979
+ 4,
1980
+ "LATENT"
1981
+ ],
1982
+ [
1983
+ 42,
1984
+ 39,
1985
+ 0,
1986
+ 33,
1987
+ 0,
1988
+ "LATENT"
1989
+ ]
1990
+ ],
1991
+ "groups": [
1992
+ {
1993
+ "id": 1,
1994
+ "title": "Models",
1995
+ "bounding": [
1996
+ -20,
1997
+ -80,
1998
+ 460,
1999
+ 480
2000
+ ],
2001
+ "color": "#3f789e",
2002
+ "font_size": 24,
2003
+ "flags": {}
2004
+ },
2005
+ {
2006
+ "id": 2,
2007
+ "title": "Prompt",
2008
+ "bounding": [
2009
+ -20,
2010
+ 440,
2011
+ 810,
2012
+ 1240
2013
+ ],
2014
+ "color": "#a1309b",
2015
+ "font_size": 24,
2016
+ "flags": {}
2017
+ },
2018
+ {
2019
+ "id": 3,
2020
+ "title": "Size: aspect ratio + 16 MP (16 GB VRAM)",
2021
+ "bounding": [
2022
+ 820,
2023
+ -80,
2024
+ 690,
2025
+ 520
2026
+ ],
2027
+ "color": "#88A",
2028
+ "font_size": 24,
2029
+ "flags": {}
2030
+ },
2031
+ {
2032
+ "id": 4,
2033
+ "title": "Reference pictures (Ctrl+B to switch a slot on)",
2034
+ "bounding": [
2035
+ 820,
2036
+ 710,
2037
+ 1470,
2038
+ 870
2039
+ ],
2040
+ "color": "#8A8",
2041
+ "font_size": 24,
2042
+ "flags": {}
2043
+ },
2044
+ {
2045
+ "id": 5,
2046
+ "title": "Render",
2047
+ "bounding": [
2048
+ 1540,
2049
+ -80,
2050
+ 1330,
2051
+ 750
2052
+ ],
2053
+ "color": "#b58b2a",
2054
+ "font_size": 24,
2055
+ "flags": {}
2056
+ }
2057
+ ],
2058
+ "config": {},
2059
+ "extra": {
2060
+ "ds": {
2061
+ "scale": 0.5,
2062
+ "offset": [
2063
+ 600,
2064
+ 150
2065
+ ]
2066
+ }
2067
+ },
2068
+ "version": 0.4
2069
+ }
Minimax_H3_image_8gb_vram.json ADDED
@@ -0,0 +1,2108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "id": "8fea2e04-7545-55d2-b4e7-a4953825a455",
3
+ "revision": 0,
4
+ "last_node_id": 60,
5
+ "last_link_id": 42,
6
+ "nodes": [
7
+ {
8
+ "properties": {
9
+ "Node name for S&R": "UNETLoader"
10
+ },
11
+ "id": 1,
12
+ "type": "UNETLoader",
13
+ "pos": [
14
+ 0,
15
+ 0
16
+ ],
17
+ "size": [
18
+ 420,
19
+ 90
20
+ ],
21
+ "flags": {},
22
+ "order": 0,
23
+ "mode": 0,
24
+ "inputs": [],
25
+ "outputs": [
26
+ {
27
+ "name": "MODEL",
28
+ "type": "MODEL",
29
+ "links": [
30
+ 33,
31
+ 36
32
+ ]
33
+ }
34
+ ],
35
+ "widgets_values": [
36
+ "minimax_h3_image_pruned_int8_convrot.safetensors",
37
+ "default"
38
+ ]
39
+ },
40
+ {
41
+ "properties": {
42
+ "Node name for S&R": "CLIPLoader"
43
+ },
44
+ "id": 2,
45
+ "type": "CLIPLoader",
46
+ "pos": [
47
+ 0,
48
+ 140
49
+ ],
50
+ "size": [
51
+ 420,
52
+ 120
53
+ ],
54
+ "flags": {},
55
+ "order": 1,
56
+ "mode": 0,
57
+ "inputs": [],
58
+ "outputs": [
59
+ {
60
+ "name": "CLIP",
61
+ "type": "CLIP",
62
+ "links": [
63
+ 15
64
+ ]
65
+ }
66
+ ],
67
+ "widgets_values": [
68
+ "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
69
+ "minimax",
70
+ "default"
71
+ ],
72
+ "title": "Load CLIP (text encoder)"
73
+ },
74
+ {
75
+ "properties": {
76
+ "Node name for S&R": "VAELoader"
77
+ },
78
+ "id": 3,
79
+ "type": "VAELoader",
80
+ "pos": [
81
+ 0,
82
+ 310
83
+ ],
84
+ "size": [
85
+ 420,
86
+ 70
87
+ ],
88
+ "flags": {},
89
+ "order": 2,
90
+ "mode": 0,
91
+ "inputs": [],
92
+ "outputs": [
93
+ {
94
+ "name": "VAE",
95
+ "type": "VAE",
96
+ "links": [
97
+ 16,
98
+ 31
99
+ ]
100
+ }
101
+ ],
102
+ "widgets_values": [
103
+ "minimax_h3_image_vae_fp16.safetensors"
104
+ ]
105
+ },
106
+ {
107
+ "properties": {
108
+ "Node name for S&R": "PrimitiveStringMultiline"
109
+ },
110
+ "id": 10,
111
+ "type": "PrimitiveStringMultiline",
112
+ "pos": [
113
+ 0,
114
+ 520
115
+ ],
116
+ "size": [
117
+ 420,
118
+ 330
119
+ ],
120
+ "flags": {},
121
+ "order": 3,
122
+ "mode": 0,
123
+ "inputs": [],
124
+ "outputs": [
125
+ {
126
+ "name": "STRING",
127
+ "type": "STRING",
128
+ "links": [
129
+ 2,
130
+ 7
131
+ ]
132
+ }
133
+ ],
134
+ "widgets_values": [
135
+ "PASTE YOUR PROMPT HERE. With reference pictures, name them as <Picture 1>, <Picture 2>, ..."
136
+ ],
137
+ "title": "PROMPT"
138
+ },
139
+ {
140
+ "properties": {
141
+ "Node name for S&R": "PrimitiveBoolean"
142
+ },
143
+ "id": 11,
144
+ "type": "PrimitiveBoolean",
145
+ "pos": [
146
+ 0,
147
+ 900
148
+ ],
149
+ "size": [
150
+ 420,
151
+ 60
152
+ ],
153
+ "flags": {},
154
+ "order": 4,
155
+ "mode": 0,
156
+ "inputs": [],
157
+ "outputs": [
158
+ {
159
+ "name": "BOOLEAN",
160
+ "type": "BOOLEAN",
161
+ "links": [
162
+ 9
163
+ ]
164
+ }
165
+ ],
166
+ "widgets_values": [
167
+ false
168
+ ],
169
+ "title": "Enhance prompt (text-to-image only)"
170
+ },
171
+ {
172
+ "properties": {
173
+ "Node name for S&R": "CLIPLoader"
174
+ },
175
+ "id": 12,
176
+ "type": "CLIPLoader",
177
+ "pos": [
178
+ 0,
179
+ 1010
180
+ ],
181
+ "size": [
182
+ 420,
183
+ 120
184
+ ],
185
+ "flags": {},
186
+ "order": 5,
187
+ "mode": 0,
188
+ "inputs": [],
189
+ "outputs": [
190
+ {
191
+ "name": "CLIP",
192
+ "type": "CLIP",
193
+ "links": [
194
+ 3
195
+ ]
196
+ }
197
+ ],
198
+ "widgets_values": [
199
+ "qwen3.5_9b_w4a8_gptq.safetensors",
200
+ "minimax",
201
+ "default"
202
+ ],
203
+ "title": "Load CLIP (prompt upsampler)"
204
+ },
205
+ {
206
+ "properties": {
207
+ "Node name for S&R": "PrimitiveStringMultiline"
208
+ },
209
+ "id": 13,
210
+ "type": "PrimitiveStringMultiline",
211
+ "pos": [
212
+ 0,
213
+ 1180
214
+ ],
215
+ "size": [
216
+ 420,
217
+ 120
218
+ ],
219
+ "flags": {},
220
+ "order": 6,
221
+ "mode": 0,
222
+ "inputs": [],
223
+ "outputs": [
224
+ {
225
+ "name": "STRING",
226
+ "type": "STRING",
227
+ "links": [
228
+ 4
229
+ ]
230
+ }
231
+ ],
232
+ "widgets_values": [
233
+ "Reason in the thinking section first. The visible final content must be exactly one JSON object with exactly two keys: expanded_caption and aspect_ratio. Every ASCII double quotation mark used as caption content inside expanded_caption must be serialized as the JSON escape sequence \\\". Do not output text outside that object."
234
+ ],
235
+ "title": "Upsampler system prompt"
236
+ },
237
+ {
238
+ "properties": {
239
+ "Node name for S&R": "PrimitiveStringMultiline"
240
+ },
241
+ "id": 14,
242
+ "type": "PrimitiveStringMultiline",
243
+ "pos": [
244
+ 0,
245
+ 1350
246
+ ],
247
+ "size": [
248
+ 420,
249
+ 120
250
+ ],
251
+ "flags": {},
252
+ "order": 7,
253
+ "mode": 0,
254
+ "inputs": [],
255
+ "outputs": [
256
+ {
257
+ "name": "STRING",
258
+ "type": "STRING",
259
+ "links": [
260
+ 1
261
+ ]
262
+ }
263
+ ],
264
+ "widgets_values": [
265
+ "You are a caption-expansion and canvas-selection model for still-image\ntext-to-image generation. Rewrite input_text into a natural English detailed\ncaption and choose one aspect ratio for the requested final image. After private\nreasoning, serialize the complete caption and aspect ratio together as the one\nrequired two-key JSON object; no visible analysis, markdown, or self-talk. The\ncaption may be long when the requested image is a complex information or design artifact.\n\nRole: act like an expert visual designer and technical illustrator, not a word\nexpander. Preserve every hard fact from the input, then complete the image with\nvisible, inspectable content. The goal is not more words; the goal is that a\nviewer can inspect each important part of the image and find meaningful subjects,\nlabels, relationships, materials, data, and hierarchy.\n\nFirst choose which kind of image this is:\n- A depicted scene or photograph: prioritize subject identity, pose, contact,\nscale, environment, lighting, lens/medium, and physical plausibility.\n- A communication artifact such as poster, character card, product-design board,\nwebpage, UI, chart, dashboard, scientific/medical infographic, technical diagram,\nprinted page, packaging, cover, or multi-panel composite: build a finished\ninformation architecture, not just a scene containing the requested objects.\n\nAspect-ratio decision policy:\nChoose exactly one of these allowed ratios: 1:1, 5:4, 4:3, 3:2, 16:9, 21:9,\n4:5, 3:4, 2:3, or 9:16. Decide in the following order; earlier evidence\noverrides later evidence.\n\n1. Explicit final canvas\n- Preserve an explicit final-output ratio. Map explicit output pixel dimensions\n by width divided by height to the closest allowed ratio.\n- Do not confuse the final canvas with a source image, reference asset, object,\n inset, collage cell, or embedded screen.\n\n2. Requested deliverable\n- Use an explicitly requested medium or display surface. Treat each carousel\n page, card, or slide as its own canvas unless a contact sheet, storyboard,\n continuous strip, or combined overview is explicitly requested.\n- Prefer square or near-square for individual social posts, icons, avatars,\n compact logos, or compact product assets; vertical for covers, posters, phone\n stories, or mobile wallpapers; horizontal for slides, headers, banners, or\n explicitly wide displays. Do not invent a platform or deliverable.\n\n3. Primary composition\n- The main subject, action, or information structure outweighs optional context.\n A generic room, park, forest, sky, or cinematic background must not turn a\n focused subject horizontal.\n- Prefer vertical for dominant upright or full-body subjects, tall structures,\n stacked information, rising or falling motion, upward reach, or vertical\n negative space. Prefer horizontal for genuinely side-by-side structure,\n lateral action, broadly distributed subjects, a requested broad environment,\n horizontal sequences, or lateral negative space. Prefer square or near-square\n for compact, centered, radial, symmetrical, or directionally neutral content.\n\n4. Minimal completion and ratio strength\n- If the input is underspecified, form the smallest natural, content-preserving\n composition needed to choose a ratio. Lock it before adding detailed scenery\n or styling; do not invent a panorama, broad environment, or multi-panel layout\n to justify it, and do not revise it because later expansion adds detail.\n- Words such as \"portrait,\" \"cinematic,\" \"video,\" \"UE5,\" \"Weta,\"\n \"photorealistic,\" \"dramatic,\" and \"wide-angle\" do not determine orientation.\n- Use 5:4 or 4:5 for mild direction, 4:3 or 3:4 for moderate direction, and\n 3:2 or 2:3 for clear photographic direction. Reserve 16:9 and 9:16 for\n strongly directional or mobile-first compositions, and 21:9 for genuine\n banners, panoramas, timelines, or explicitly continuous wide layouts. When\n evidence remains weak, prefer 1:1 or a moderate ratio.\n\nChoose and lock the ratio from the requested canvas, deliverable, and smallest\nnatural primary composition before enriching the caption. Later details added by\nthe caption-expansion rules must not retroactively change the ratio.\n\nUniversal hard rules:\n1. Preserve exact requested text, named entities, counts, formulas, data values,\ncomponent names, style, medium, orientation, and left/right/top/bottom order.\n2. Required content must be visible, not implied. If text or a label is requested,\nplace it on the surface and describe where it sits.\n3. For active bodies, solve scene geometry before pose detail. Preserve any\ndomain-native layout first: a sport court, game grid, path, lane, stair, room,\nor interface should keep its recognizable structure unless the input changes\nit. Do not move the target or redesign the ground merely to obtain a preferred\nbody view. Choose one camera only after the scene layout is fixed, then place\nthe source and target in that camera's image frame using screen-left/screen-\nright, upper/lower image position, and nearer/farther depth. Silently trace the\nsource-to-target vector in this fixed view. Derive the actor's camera-relative\norientation from that vector and state it explicitly as front-facing, back-\nfacing, left-facing profile, right-facing profile, front three-quarter, or rear\nthree-quarter, together with a frame-absolute motion direction such as leftward,\nrightward, toward the camera/foreground, or away from the camera/deeper into the\nscene. Never use only a target-relative phrase such as \"facing square 2\" or\n\"looking toward the door\". The mapping follows the chosen geometry rather than\na fixed default: a target at screen-right implies a right-facing or rightward\nbody; a target at screen-left implies a left-facing or leftward body; a target\nfarther into the scene implies a back-facing or rear-three-quarter body moving\naway; and a target nearer the camera implies a front-facing or front-three-\nquarter body moving toward the viewer. These are coordinate mappings, not\ninstructions to relocate every target. Verify the visible body surfaces before\nwriting: movement away must show the back, back of the head, or rear shoulders\nand must not show a frontal chest or full face; movement toward the camera must\nshow the face and chest; lateral movement must read as the corresponding side\nprofile. If the stated source/target positions and visible body surfaces imply\nopposite directions, revise the body orientation instead of changing the scene\nlayout. Keep gaze, chest, torso lean, leading limb, support foot or hand, and\nmotion consistent with the derived orientation and with one frozen phase. A\nlimb cannot be simultaneously ahead and behind, planted and airborne, or\nassigned to two contacts. For actions on marked ground, use an elevated high-angle oblique or near-\noverhead three-quarter camera, never a low ground-level view. Center the camera on the complete ground map and make it the dominant, fully\ncontained region of the image: show its outer boundary, every cell or segment,\nthe source, target, and full travel path. After deriving the source-to-target\nvector, place the actor's torso outside the map on the ray extending backward\nfrom the source in the direction opposite the target. Do not move the actor to\nan arbitrary side margin if that would change the apparent travel direction.\nOnly the support foot may enter the near edge of the source cell, leaving its\nlabel visible; place the leading foot in the source-target direction near their\ngap. Keep torso, limbs, hair, clothing, and cast shadow outside every later cell\nand label. Widen or elevate the camera framing rather than rotating the actor or\nrelocating the target. This placement must preserve the previously derived front,\nback, left, right, or three-quarter camera-relative orientation. Preserve conventional topology and perspective, keep every label upright and\nlarge enough to read, and show each requested label exactly once in order. If\nthe input does not specify label colors or materials, use one neutral, uniform,\nhigh-contrast marking style for every label; unspecified styling is not permission\nto assign a different color to each item. Keep the actor's hands empty unless the\ninput names a held object. Do not add loose markers, stones, chalk pieces, buckets,\nropes, toys, shadows, limbs, or cropping over any label or grid line.\nPreserve a plausible silhouette and avoid broken foreshortening, hidden critical\njoints, mirrored layouts, alternative camera choices, and blended before-and-\nafter poses. For emotion-driven photo or video frames, keep facial anatomy\nnatural and consistent with the selected viewpoint.\n4. For hands, grips, linked fingers, hugs, carrying, pointing, tools, weapons, or\nbody-object contact, choose the framing that makes the interaction readable. Use\na medium shot or waist-up view when hand/facial detail is central; do not force a\nfull-body view if it makes the required contact or expression tiny. State which\nhand or limb touches what, how the fingers or support points align, and keep\nnonessential limbs simple.\n5. Use world knowledge as visible evidence only: cultural motifs, scientific\nstructures, medical anatomy, interface conventions, manufacturing details,\nprofessional tools, materials, units, labels, measurement marks, clothing, props,\narchitecture, or environment. Do not add lore that cannot be seen.\n6. Respect requested subject count. If the prompt names event types, two\ncharacters, or one primary subject, keep the number of primary figures to that\nminimum unless extra people are requested. Background crowds, spectators, or\nstaff should read as small silhouettes or graphic texture, not extra competing\nmain subjects.\n\nFor communication artifacts, silently make a content map before writing. Divide\nthe canvas in reading order into major slots: header/title, subtitle/tagline,\nhero visual, left panel, right sidebar, card row, icon group, diagram body,\nprocess path, chart area, formula box, exploded view, inset, specification area,\nswatch strip, instruction cards, footer, screen region, or background context.\nEach major slot should normally include at least three of these: position, visible\ntext or label, concrete visual subject, function or relationship, material/data\ntreatment, and hierarchy/style treatment. Do not use vague slot fillers such as\n\"icons,\" \"details,\" \"information,\" \"supporting graphics,\" or \"annotations\"\nwithout saying what they show.\n\nMake the artifact feel complete. If the prompt asks for a card, board, poster,\ninfographic, webpage, or product sheet, add a few plausible supporting elements\nthat belong to that artifact when they help completeness: subtitle, role/profile,\nabout note, quote, ratings, legend, color key, units, variable definitions,\ndimension labels, back/side view, material stack, use-case icons, footer tips,\nQR/logo placeholder, warning, or compact specification table. Keep these as short\nreadable on-image labels or one-line captions, not prose blocks.\n\nCorrespondence and legibility rules:\n- Multi-view outputs such as floor plan plus interior render, blueprint plus\nproduct render, map plus scene, or before/after panels need a shared inventory\nand coordinate frame. Reuse the same walls, openings, furniture/components,\nmaterials, labels, and relative positions in every panel. Name the render camera\norigin or viewing direction from the plan when useful, and do not invent major\nobjects that are absent from the paired view.\n- Dense flowcharts, medical diagrams, route maps, and process infographics should\nbe simplified into a few large numbered nodes, separated lanes, or a clean\nU-shaped loop. Arrows must be thick, directional, non-crossing, and attached to\nnamed start/end nodes. Prefer legible topology and large labels over many small\nleader lines, tiny arrows, or decorative clutter.\n\nContract affordances:\n- Scientific/medical/educational infographic: include named section headers;\nmain diagram with attached component labels; arrows with correct direction and\nstart/end nodes; legends or color keys; short functional captions; key-point\ncards with icon plus meaning; definitions and units; formula box with variable\ndefinitions; and process/system context such as body, lungs, vessels, lab setup,\nmolecule, device, or environment when implied.\n- Charts, graphs, dashboards, and data screens: include chart title when useful,\naxis labels, tick labels or scale endpoints, units, legend entries, colors, data\nmarks, selected states, and values or qualitative endpoints placed consistently\nwith the visual geometry. Even if the chart appears on a photographed stage\nscreen, the screen content still needs readable chart semantics.\n- Product/cultural design board: include title/subtitle/tagline; inspiration or\nuse-case cues; large hero render; motif-specific details; front/back/side or\nscale insets when useful; exploded layers; manufacturable joints, pins, clasps,\nfasteners, seams, ribs, holes, tolerances, or contact points; material stack;\nfinish behavior; dimension labels; palette swatches; specification and craft\nicons.\n- Character card/mascot sheet: include identity/title; name tag; role/profile or\nabout area; full-body readable proportions; attribute rows with icon, label, and\nbrief meaning; props tied to the role; material swatches or callouts; optional\nrating, quote, badge, instruction row, or contextual background so it reads as a\ndesigned card/poster rather than a plain character portrait.\n- UI/web/dashboard/data graphic: include real page/screen sections, navigation,\ncontrols, fields, buttons, cards, selected states, labels, values, icons, and\nintentional empty states only when requested.\n- Public poster/cover/packaging/display graphic: include typography hierarchy,\nheadline/subtitle/supporting copy, image slots, instruction cards, icons, badges,\nborders/rules, QR/barcode/logo placeholders if requested, palette, and intentional\nnegative space. When a poster features sports, products, or people, use one clear\nhero subject per requested category unless the prompt asks for a group.\n\nStyle comes after content. Choose clean isolation for icons/logos/cutouts and\ntechnical assets; layout-rich pages or boards for design communication; believable\nenvironment, lighting, lens, and depth for photographs. Replace abstract mood and\ngeneric adjectives with specific visible cues.\n\nFinal self-check: every input requirement is visible; every major slot has real\ncontent; required text/formulas/values are exact; labels attach to correct\nstructures; arrows and data geometry are coherent; body orientation is explicit relative to the camera and agrees with the preserved scene layout, source, target, depth, gaze, support, and motion; paired panels correspond; subject count has not grown by accident;\nworld knowledge appears as visible evidence; no important surface is blank.\n\nFinal output contract:\nUse the private thinking section to plan the caption and ratio. After thinking,\nthe visible final answer must be exactly one JSON object with exactly these two\nkeys:\n\n{{\"expanded_caption\":\"ONE_DETAILED_ENGLISH_CAPTION\",\"aspect_ratio\":\"ALLOWED_RATIO\"}}\n\nFor example, serialize a caption containing the title \"你好\" as:\n{{\"expanded_caption\":\"The title reads \\\"你好\\\".\",\"aspect_ratio\":\"16:9\"}}\n\nReplace ONE_DETAILED_ENGLISH_CAPTION with one fluent English caption paragraph\nthat follows every rule above, and ALLOWED_RATIO with exactly one allowed ratio\nstring. Keep expanded_caption as one valid JSON string. Every ASCII double quotation mark used as caption content inside expanded_caption must be serialized as the JSON escape sequence \\\". Escape all other characters as valid JSON requires. Do not add or omit keys. Do not put thinking or think tags inside\nthe object. Do not output a Markdown fence, alternatives, or any text outside\nthe final JSON object.\n\ninput_text: {input text}\noutput_text:\n"
266
+ ],
267
+ "title": "Upsampler task prompt"
268
+ },
269
+ {
270
+ "properties": {
271
+ "Node name for S&R": "StringReplace"
272
+ },
273
+ "id": 15,
274
+ "type": "StringReplace",
275
+ "pos": [
276
+ 470,
277
+ 900
278
+ ],
279
+ "size": [
280
+ 300,
281
+ 200
282
+ ],
283
+ "flags": {},
284
+ "order": 8,
285
+ "mode": 0,
286
+ "inputs": [
287
+ {
288
+ "name": "string",
289
+ "type": "STRING",
290
+ "link": 1,
291
+ "widget": {
292
+ "name": "string"
293
+ }
294
+ },
295
+ {
296
+ "name": "find",
297
+ "type": "STRING",
298
+ "link": null,
299
+ "widget": {
300
+ "name": "find"
301
+ }
302
+ },
303
+ {
304
+ "name": "replace",
305
+ "type": "STRING",
306
+ "link": 2,
307
+ "widget": {
308
+ "name": "replace"
309
+ }
310
+ }
311
+ ],
312
+ "outputs": [
313
+ {
314
+ "name": "STRING",
315
+ "type": "STRING",
316
+ "links": [
317
+ 5
318
+ ]
319
+ }
320
+ ],
321
+ "widgets_values": [
322
+ "",
323
+ "{input text}",
324
+ ""
325
+ ]
326
+ },
327
+ {
328
+ "properties": {
329
+ "Node name for S&R": "TextGenerateV2"
330
+ },
331
+ "id": 16,
332
+ "type": "TextGenerateV2",
333
+ "pos": [
334
+ 470,
335
+ 1160
336
+ ],
337
+ "size": [
338
+ 300,
339
+ 460
340
+ ],
341
+ "flags": {},
342
+ "order": 9,
343
+ "mode": 0,
344
+ "inputs": [
345
+ {
346
+ "name": "clip",
347
+ "type": "CLIP",
348
+ "link": 3
349
+ },
350
+ {
351
+ "name": "system_prompt",
352
+ "type": "STRING",
353
+ "link": 4
354
+ },
355
+ {
356
+ "name": "prompt",
357
+ "type": "STRING",
358
+ "link": 5,
359
+ "widget": {
360
+ "name": "prompt"
361
+ }
362
+ }
363
+ ],
364
+ "outputs": [
365
+ {
366
+ "name": "generated_text",
367
+ "type": "STRING",
368
+ "links": [
369
+ 6
370
+ ]
371
+ },
372
+ {
373
+ "name": "thinking",
374
+ "type": "STRING",
375
+ "links": []
376
+ }
377
+ ],
378
+ "widgets_values": [
379
+ "",
380
+ 8192,
381
+ "on",
382
+ 1,
383
+ 20,
384
+ 0.95,
385
+ 0,
386
+ 1,
387
+ 0,
388
+ 0,
389
+ true,
390
+ true,
391
+ "auto"
392
+ ]
393
+ },
394
+ {
395
+ "properties": {
396
+ "Node name for S&R": "JsonExtractString"
397
+ },
398
+ "id": 17,
399
+ "type": "JsonExtractString",
400
+ "pos": [
401
+ 0,
402
+ 1520
403
+ ],
404
+ "size": [
405
+ 420,
406
+ 110
407
+ ],
408
+ "flags": {},
409
+ "order": 10,
410
+ "mode": 0,
411
+ "inputs": [
412
+ {
413
+ "name": "json_string",
414
+ "type": "STRING",
415
+ "link": 6,
416
+ "widget": {
417
+ "name": "json_string"
418
+ }
419
+ },
420
+ {
421
+ "name": "key",
422
+ "type": "STRING",
423
+ "link": null,
424
+ "widget": {
425
+ "name": "key"
426
+ }
427
+ }
428
+ ],
429
+ "outputs": [
430
+ {
431
+ "name": "STRING",
432
+ "type": "STRING",
433
+ "links": [
434
+ 8
435
+ ]
436
+ }
437
+ ],
438
+ "widgets_values": [
439
+ "",
440
+ "expanded_caption"
441
+ ]
442
+ },
443
+ {
444
+ "properties": {
445
+ "Node name for S&R": "ComfySwitchNode"
446
+ },
447
+ "id": 18,
448
+ "type": "ComfySwitchNode",
449
+ "pos": [
450
+ 470,
451
+ 520
452
+ ],
453
+ "size": [
454
+ 300,
455
+ 90
456
+ ],
457
+ "flags": {},
458
+ "order": 11,
459
+ "mode": 0,
460
+ "inputs": [
461
+ {
462
+ "name": "on_false",
463
+ "type": "STRING",
464
+ "link": 7
465
+ },
466
+ {
467
+ "name": "on_true",
468
+ "type": "STRING",
469
+ "link": 8
470
+ },
471
+ {
472
+ "name": "switch",
473
+ "type": "BOOLEAN",
474
+ "link": 9,
475
+ "widget": {
476
+ "name": "switch"
477
+ }
478
+ }
479
+ ],
480
+ "outputs": [
481
+ {
482
+ "name": "output",
483
+ "type": "STRING",
484
+ "links": [
485
+ 10,
486
+ 27
487
+ ]
488
+ }
489
+ ],
490
+ "widgets_values": [
491
+ false
492
+ ],
493
+ "title": "Prompt: as written / enhanced"
494
+ },
495
+ {
496
+ "properties": {
497
+ "Node name for S&R": "PreviewAny"
498
+ },
499
+ "id": 19,
500
+ "type": "PreviewAny",
501
+ "pos": [
502
+ 470,
503
+ 660
504
+ ],
505
+ "size": [
506
+ 300,
507
+ 190
508
+ ],
509
+ "flags": {},
510
+ "order": 12,
511
+ "mode": 0,
512
+ "inputs": [
513
+ {
514
+ "name": "source",
515
+ "type": "*",
516
+ "link": 10
517
+ }
518
+ ],
519
+ "outputs": [
520
+ {
521
+ "name": "STRING",
522
+ "type": "STRING",
523
+ "links": []
524
+ }
525
+ ],
526
+ "widgets_values": [],
527
+ "title": "Prompt sent to the model"
528
+ },
529
+ {
530
+ "properties": {
531
+ "Node name for S&R": "PrimitiveBoolean"
532
+ },
533
+ "id": 20,
534
+ "type": "PrimitiveBoolean",
535
+ "pos": [
536
+ 840,
537
+ 0
538
+ ],
539
+ "size": [
540
+ 300,
541
+ 60
542
+ ],
543
+ "flags": {},
544
+ "order": 13,
545
+ "mode": 0,
546
+ "inputs": [],
547
+ "outputs": [
548
+ {
549
+ "name": "BOOLEAN",
550
+ "type": "BOOLEAN",
551
+ "links": [
552
+ 13
553
+ ]
554
+ }
555
+ ],
556
+ "widgets_values": [
557
+ true
558
+ ],
559
+ "title": "16 MP (off = 4 MP draft)"
560
+ },
561
+ {
562
+ "properties": {
563
+ "Node name for S&R": "PrimitiveFloat"
564
+ },
565
+ "id": 21,
566
+ "type": "PrimitiveFloat",
567
+ "pos": [
568
+ 840,
569
+ 110
570
+ ],
571
+ "size": [
572
+ 300,
573
+ 60
574
+ ],
575
+ "flags": {},
576
+ "order": 14,
577
+ "mode": 0,
578
+ "inputs": [],
579
+ "outputs": [
580
+ {
581
+ "name": "FLOAT",
582
+ "type": "FLOAT",
583
+ "links": [
584
+ 11
585
+ ]
586
+ }
587
+ ],
588
+ "widgets_values": [
589
+ 4.0
590
+ ],
591
+ "title": "Draft megapixels"
592
+ },
593
+ {
594
+ "properties": {
595
+ "Node name for S&R": "PrimitiveFloat"
596
+ },
597
+ "id": 22,
598
+ "type": "PrimitiveFloat",
599
+ "pos": [
600
+ 840,
601
+ 220
602
+ ],
603
+ "size": [
604
+ 300,
605
+ 60
606
+ ],
607
+ "flags": {},
608
+ "order": 15,
609
+ "mode": 0,
610
+ "inputs": [],
611
+ "outputs": [
612
+ {
613
+ "name": "FLOAT",
614
+ "type": "FLOAT",
615
+ "links": [
616
+ 12
617
+ ]
618
+ }
619
+ ],
620
+ "widgets_values": [
621
+ 16.0
622
+ ],
623
+ "title": "Full megapixels (max 16)"
624
+ },
625
+ {
626
+ "properties": {
627
+ "Node name for S&R": "ComfySwitchNode"
628
+ },
629
+ "id": 23,
630
+ "type": "ComfySwitchNode",
631
+ "pos": [
632
+ 840,
633
+ 330
634
+ ],
635
+ "size": [
636
+ 300,
637
+ 90
638
+ ],
639
+ "flags": {},
640
+ "order": 16,
641
+ "mode": 0,
642
+ "inputs": [
643
+ {
644
+ "name": "on_false",
645
+ "type": "FLOAT",
646
+ "link": 11
647
+ },
648
+ {
649
+ "name": "on_true",
650
+ "type": "FLOAT",
651
+ "link": 12
652
+ },
653
+ {
654
+ "name": "switch",
655
+ "type": "BOOLEAN",
656
+ "link": 13,
657
+ "widget": {
658
+ "name": "switch"
659
+ }
660
+ }
661
+ ],
662
+ "outputs": [
663
+ {
664
+ "name": "output",
665
+ "type": "FLOAT",
666
+ "links": [
667
+ 14
668
+ ]
669
+ }
670
+ ],
671
+ "widgets_values": [
672
+ false
673
+ ],
674
+ "title": "Megapixels: draft / 16 MP"
675
+ },
676
+ {
677
+ "properties": {
678
+ "Node name for S&R": "ResolutionSelector"
679
+ },
680
+ "id": 24,
681
+ "type": "ResolutionSelector",
682
+ "pos": [
683
+ 1190,
684
+ 0
685
+ ],
686
+ "size": [
687
+ 300,
688
+ 240
689
+ ],
690
+ "flags": {},
691
+ "order": 17,
692
+ "mode": 0,
693
+ "inputs": [
694
+ {
695
+ "name": "aspect_ratio",
696
+ "type": "COMBO",
697
+ "link": null,
698
+ "widget": {
699
+ "name": "aspect_ratio"
700
+ }
701
+ },
702
+ {
703
+ "name": "megapixels",
704
+ "type": "FLOAT",
705
+ "link": 14,
706
+ "widget": {
707
+ "name": "megapixels"
708
+ }
709
+ },
710
+ {
711
+ "name": "multiple",
712
+ "type": "INT",
713
+ "link": null,
714
+ "widget": {
715
+ "name": "multiple"
716
+ }
717
+ }
718
+ ],
719
+ "outputs": [
720
+ {
721
+ "name": "width",
722
+ "type": "INT",
723
+ "links": [
724
+ 28
725
+ ]
726
+ },
727
+ {
728
+ "name": "height",
729
+ "type": "INT",
730
+ "links": [
731
+ 29
732
+ ]
733
+ }
734
+ ],
735
+ "widgets_values": [
736
+ "16:9 (Widescreen)",
737
+ 16.0,
738
+ 32
739
+ ],
740
+ "title": "ASPECT RATIO"
741
+ },
742
+ {
743
+ "properties": {
744
+ "Node name for S&R": "MiniMaxH3ImageConditioning"
745
+ },
746
+ "id": 30,
747
+ "type": "MiniMaxH3ImageConditioning",
748
+ "pos": [
749
+ 1560,
750
+ 0
751
+ ],
752
+ "size": [
753
+ 330,
754
+ 400
755
+ ],
756
+ "flags": {},
757
+ "order": 18,
758
+ "mode": 0,
759
+ "inputs": [
760
+ {
761
+ "name": "clip",
762
+ "type": "CLIP",
763
+ "link": 15
764
+ },
765
+ {
766
+ "name": "vae",
767
+ "type": "VAE",
768
+ "link": 16
769
+ },
770
+ {
771
+ "name": "images.image_1",
772
+ "type": "IMAGE",
773
+ "link": 17
774
+ },
775
+ {
776
+ "name": "images.image_2",
777
+ "type": "IMAGE",
778
+ "link": 18
779
+ },
780
+ {
781
+ "name": "images.image_3",
782
+ "type": "IMAGE",
783
+ "link": 19
784
+ },
785
+ {
786
+ "name": "images.image_4",
787
+ "type": "IMAGE",
788
+ "link": 20
789
+ },
790
+ {
791
+ "name": "images.image_5",
792
+ "type": "IMAGE",
793
+ "link": 21
794
+ },
795
+ {
796
+ "name": "images.image_6",
797
+ "type": "IMAGE",
798
+ "link": 22
799
+ },
800
+ {
801
+ "name": "images.image_7",
802
+ "type": "IMAGE",
803
+ "link": 23
804
+ },
805
+ {
806
+ "name": "images.image_8",
807
+ "type": "IMAGE",
808
+ "link": 24
809
+ },
810
+ {
811
+ "name": "images.image_9",
812
+ "type": "IMAGE",
813
+ "link": 25
814
+ },
815
+ {
816
+ "name": "images.image_10",
817
+ "type": "IMAGE",
818
+ "link": 26
819
+ },
820
+ {
821
+ "name": "prompt",
822
+ "type": "STRING",
823
+ "link": 27,
824
+ "widget": {
825
+ "name": "prompt"
826
+ }
827
+ },
828
+ {
829
+ "name": "width",
830
+ "type": "INT",
831
+ "link": 28,
832
+ "widget": {
833
+ "name": "width"
834
+ }
835
+ },
836
+ {
837
+ "name": "height",
838
+ "type": "INT",
839
+ "link": 29,
840
+ "widget": {
841
+ "name": "height"
842
+ }
843
+ }
844
+ ],
845
+ "outputs": [
846
+ {
847
+ "name": "positive",
848
+ "type": "CONDITIONING",
849
+ "links": [
850
+ 30,
851
+ 34
852
+ ]
853
+ },
854
+ {
855
+ "name": "LATENT",
856
+ "type": "LATENT",
857
+ "links": [
858
+ 41
859
+ ]
860
+ }
861
+ ],
862
+ "widgets_values": [
863
+ "",
864
+ 0,
865
+ 0
866
+ ]
867
+ },
868
+ {
869
+ "properties": {
870
+ "Node name for S&R": "ConditioningZeroOut"
871
+ },
872
+ "id": 31,
873
+ "type": "ConditioningZeroOut",
874
+ "pos": [
875
+ 1940,
876
+ 0
877
+ ],
878
+ "size": [
879
+ 300,
880
+ 50
881
+ ],
882
+ "flags": {},
883
+ "order": 19,
884
+ "mode": 0,
885
+ "inputs": [
886
+ {
887
+ "name": "conditioning",
888
+ "type": "CONDITIONING",
889
+ "link": 30
890
+ }
891
+ ],
892
+ "outputs": [
893
+ {
894
+ "name": "CONDITIONING",
895
+ "type": "CONDITIONING",
896
+ "links": [
897
+ 35
898
+ ]
899
+ }
900
+ ]
901
+ },
902
+ {
903
+ "properties": {
904
+ "Node name for S&R": "VAEDecodeTiled"
905
+ },
906
+ "id": 33,
907
+ "type": "VAEDecodeTiled",
908
+ "pos": [
909
+ 2290,
910
+ 0
911
+ ],
912
+ "size": [
913
+ 260,
914
+ 150
915
+ ],
916
+ "flags": {},
917
+ "order": 20,
918
+ "mode": 0,
919
+ "inputs": [
920
+ {
921
+ "name": "samples",
922
+ "type": "LATENT",
923
+ "link": 42
924
+ },
925
+ {
926
+ "name": "vae",
927
+ "type": "VAE",
928
+ "link": 31
929
+ },
930
+ {
931
+ "name": "tile_size",
932
+ "type": "INT",
933
+ "link": null,
934
+ "widget": {
935
+ "name": "tile_size"
936
+ }
937
+ },
938
+ {
939
+ "name": "overlap",
940
+ "type": "INT",
941
+ "link": null,
942
+ "widget": {
943
+ "name": "overlap"
944
+ }
945
+ },
946
+ {
947
+ "name": "temporal_size",
948
+ "type": "INT",
949
+ "link": null,
950
+ "widget": {
951
+ "name": "temporal_size"
952
+ }
953
+ },
954
+ {
955
+ "name": "temporal_overlap",
956
+ "type": "INT",
957
+ "link": null,
958
+ "widget": {
959
+ "name": "temporal_overlap"
960
+ }
961
+ }
962
+ ],
963
+ "outputs": [
964
+ {
965
+ "name": "IMAGE",
966
+ "type": "IMAGE",
967
+ "links": [
968
+ 32
969
+ ]
970
+ }
971
+ ],
972
+ "title": "VAE Decode (tiled, low VRAM)",
973
+ "widgets_values": [
974
+ 512,
975
+ 64,
976
+ 64,
977
+ 8
978
+ ]
979
+ },
980
+ {
981
+ "properties": {
982
+ "Node name for S&R": "SaveImage"
983
+ },
984
+ "id": 34,
985
+ "type": "SaveImage",
986
+ "pos": [
987
+ 2290,
988
+ 200
989
+ ],
990
+ "size": [
991
+ 560,
992
+ 520
993
+ ],
994
+ "flags": {},
995
+ "order": 21,
996
+ "mode": 0,
997
+ "inputs": [
998
+ {
999
+ "name": "images",
1000
+ "type": "IMAGE",
1001
+ "link": 32
1002
+ }
1003
+ ],
1004
+ "outputs": [],
1005
+ "widgets_values": [
1006
+ "h3_image/minimax_h3_image_8gb"
1007
+ ]
1008
+ },
1009
+ {
1010
+ "properties": {
1011
+ "Node name for S&R": "RandomNoise"
1012
+ },
1013
+ "id": 35,
1014
+ "type": "RandomNoise",
1015
+ "pos": [
1016
+ 1940,
1017
+ 90
1018
+ ],
1019
+ "size": [
1020
+ 300,
1021
+ 82
1022
+ ],
1023
+ "flags": {},
1024
+ "order": 22,
1025
+ "mode": 0,
1026
+ "inputs": [
1027
+ {
1028
+ "name": "noise_seed",
1029
+ "type": "INT",
1030
+ "link": null,
1031
+ "widget": {
1032
+ "name": "noise_seed"
1033
+ }
1034
+ }
1035
+ ],
1036
+ "outputs": [
1037
+ {
1038
+ "name": "NOISE",
1039
+ "type": "NOISE",
1040
+ "links": [
1041
+ 37
1042
+ ]
1043
+ }
1044
+ ],
1045
+ "widgets_values": [
1046
+ 42,
1047
+ "randomize"
1048
+ ],
1049
+ "title": "Seed"
1050
+ },
1051
+ {
1052
+ "properties": {
1053
+ "Node name for S&R": "CFGGuider"
1054
+ },
1055
+ "id": 36,
1056
+ "type": "CFGGuider",
1057
+ "pos": [
1058
+ 1940,
1059
+ 200
1060
+ ],
1061
+ "size": [
1062
+ 300,
1063
+ 100
1064
+ ],
1065
+ "flags": {},
1066
+ "order": 23,
1067
+ "mode": 0,
1068
+ "inputs": [
1069
+ {
1070
+ "name": "model",
1071
+ "type": "MODEL",
1072
+ "link": 33
1073
+ },
1074
+ {
1075
+ "name": "positive",
1076
+ "type": "CONDITIONING",
1077
+ "link": 34
1078
+ },
1079
+ {
1080
+ "name": "negative",
1081
+ "type": "CONDITIONING",
1082
+ "link": 35
1083
+ },
1084
+ {
1085
+ "name": "cfg",
1086
+ "type": "FLOAT",
1087
+ "link": null,
1088
+ "widget": {
1089
+ "name": "cfg"
1090
+ }
1091
+ }
1092
+ ],
1093
+ "outputs": [
1094
+ {
1095
+ "name": "GUIDER",
1096
+ "type": "GUIDER",
1097
+ "links": [
1098
+ 38
1099
+ ]
1100
+ }
1101
+ ],
1102
+ "widgets_values": [
1103
+ 1.0
1104
+ ]
1105
+ },
1106
+ {
1107
+ "properties": {
1108
+ "Node name for S&R": "BasicScheduler"
1109
+ },
1110
+ "id": 37,
1111
+ "type": "BasicScheduler",
1112
+ "pos": [
1113
+ 1940,
1114
+ 330
1115
+ ],
1116
+ "size": [
1117
+ 300,
1118
+ 110
1119
+ ],
1120
+ "flags": {},
1121
+ "order": 24,
1122
+ "mode": 0,
1123
+ "inputs": [
1124
+ {
1125
+ "name": "model",
1126
+ "type": "MODEL",
1127
+ "link": 36
1128
+ },
1129
+ {
1130
+ "name": "scheduler",
1131
+ "type": "COMBO",
1132
+ "link": null,
1133
+ "widget": {
1134
+ "name": "scheduler"
1135
+ }
1136
+ },
1137
+ {
1138
+ "name": "steps",
1139
+ "type": "INT",
1140
+ "link": null,
1141
+ "widget": {
1142
+ "name": "steps"
1143
+ }
1144
+ },
1145
+ {
1146
+ "name": "denoise",
1147
+ "type": "FLOAT",
1148
+ "link": null,
1149
+ "widget": {
1150
+ "name": "denoise"
1151
+ }
1152
+ }
1153
+ ],
1154
+ "outputs": [
1155
+ {
1156
+ "name": "SIGMAS",
1157
+ "type": "SIGMAS",
1158
+ "links": [
1159
+ 40
1160
+ ]
1161
+ }
1162
+ ],
1163
+ "widgets_values": [
1164
+ "simple",
1165
+ 6,
1166
+ 1.0
1167
+ ]
1168
+ },
1169
+ {
1170
+ "properties": {
1171
+ "Node name for S&R": "SamplerLCM"
1172
+ },
1173
+ "id": 38,
1174
+ "type": "SamplerLCM",
1175
+ "pos": [
1176
+ 1940,
1177
+ 470
1178
+ ],
1179
+ "size": [
1180
+ 300,
1181
+ 110
1182
+ ],
1183
+ "flags": {},
1184
+ "order": 25,
1185
+ "mode": 0,
1186
+ "inputs": [
1187
+ {
1188
+ "name": "s_noise",
1189
+ "type": "FLOAT",
1190
+ "link": null,
1191
+ "widget": {
1192
+ "name": "s_noise"
1193
+ }
1194
+ },
1195
+ {
1196
+ "name": "s_noise_end",
1197
+ "type": "FLOAT",
1198
+ "link": null,
1199
+ "widget": {
1200
+ "name": "s_noise_end"
1201
+ }
1202
+ },
1203
+ {
1204
+ "name": "noise_clip_std",
1205
+ "type": "FLOAT",
1206
+ "link": null,
1207
+ "widget": {
1208
+ "name": "noise_clip_std"
1209
+ }
1210
+ }
1211
+ ],
1212
+ "outputs": [
1213
+ {
1214
+ "name": "SAMPLER",
1215
+ "type": "SAMPLER",
1216
+ "links": [
1217
+ 39
1218
+ ]
1219
+ }
1220
+ ],
1221
+ "widgets_values": [
1222
+ 1.1,
1223
+ 0.9,
1224
+ 0.0
1225
+ ],
1226
+ "title": "SamplerLCM (real skin)"
1227
+ },
1228
+ {
1229
+ "properties": {
1230
+ "Node name for S&R": "SamplerCustomAdvanced"
1231
+ },
1232
+ "id": 39,
1233
+ "type": "SamplerCustomAdvanced",
1234
+ "pos": [
1235
+ 1560,
1236
+ 440
1237
+ ],
1238
+ "size": [
1239
+ 330,
1240
+ 130
1241
+ ],
1242
+ "flags": {},
1243
+ "order": 26,
1244
+ "mode": 0,
1245
+ "inputs": [
1246
+ {
1247
+ "name": "noise",
1248
+ "type": "NOISE",
1249
+ "link": 37
1250
+ },
1251
+ {
1252
+ "name": "guider",
1253
+ "type": "GUIDER",
1254
+ "link": 38
1255
+ },
1256
+ {
1257
+ "name": "sampler",
1258
+ "type": "SAMPLER",
1259
+ "link": 39
1260
+ },
1261
+ {
1262
+ "name": "sigmas",
1263
+ "type": "SIGMAS",
1264
+ "link": 40
1265
+ },
1266
+ {
1267
+ "name": "latent_image",
1268
+ "type": "LATENT",
1269
+ "link": 41
1270
+ }
1271
+ ],
1272
+ "outputs": [
1273
+ {
1274
+ "name": "output",
1275
+ "type": "LATENT",
1276
+ "links": [
1277
+ 42
1278
+ ]
1279
+ },
1280
+ {
1281
+ "name": "denoised_output",
1282
+ "type": "LATENT",
1283
+ "links": []
1284
+ }
1285
+ ]
1286
+ },
1287
+ {
1288
+ "properties": {
1289
+ "Node name for S&R": "LoadImage"
1290
+ },
1291
+ "id": 41,
1292
+ "type": "LoadImage",
1293
+ "pos": [
1294
+ 840,
1295
+ 790
1296
+ ],
1297
+ "size": [
1298
+ 270,
1299
+ 360
1300
+ ],
1301
+ "flags": {},
1302
+ "order": 27,
1303
+ "mode": 4,
1304
+ "inputs": [],
1305
+ "outputs": [
1306
+ {
1307
+ "name": "IMAGE",
1308
+ "type": "IMAGE",
1309
+ "links": [
1310
+ 17
1311
+ ]
1312
+ },
1313
+ {
1314
+ "name": "MASK",
1315
+ "type": "MASK",
1316
+ "links": []
1317
+ }
1318
+ ],
1319
+ "widgets_values": [
1320
+ "01_Elena.png",
1321
+ "image"
1322
+ ],
1323
+ "title": "Picture 1 (the one to edit)"
1324
+ },
1325
+ {
1326
+ "properties": {
1327
+ "Node name for S&R": "LoadImage"
1328
+ },
1329
+ "id": 42,
1330
+ "type": "LoadImage",
1331
+ "pos": [
1332
+ 1130,
1333
+ 790
1334
+ ],
1335
+ "size": [
1336
+ 270,
1337
+ 360
1338
+ ],
1339
+ "flags": {},
1340
+ "order": 28,
1341
+ "mode": 4,
1342
+ "inputs": [],
1343
+ "outputs": [
1344
+ {
1345
+ "name": "IMAGE",
1346
+ "type": "IMAGE",
1347
+ "links": [
1348
+ 18
1349
+ ]
1350
+ },
1351
+ {
1352
+ "name": "MASK",
1353
+ "type": "MASK",
1354
+ "links": []
1355
+ }
1356
+ ],
1357
+ "widgets_values": [
1358
+ "01_Elena.png",
1359
+ "image"
1360
+ ],
1361
+ "title": "Picture 2"
1362
+ },
1363
+ {
1364
+ "properties": {
1365
+ "Node name for S&R": "LoadImage"
1366
+ },
1367
+ "id": 43,
1368
+ "type": "LoadImage",
1369
+ "pos": [
1370
+ 1420,
1371
+ 790
1372
+ ],
1373
+ "size": [
1374
+ 270,
1375
+ 360
1376
+ ],
1377
+ "flags": {},
1378
+ "order": 29,
1379
+ "mode": 4,
1380
+ "inputs": [],
1381
+ "outputs": [
1382
+ {
1383
+ "name": "IMAGE",
1384
+ "type": "IMAGE",
1385
+ "links": [
1386
+ 19
1387
+ ]
1388
+ },
1389
+ {
1390
+ "name": "MASK",
1391
+ "type": "MASK",
1392
+ "links": []
1393
+ }
1394
+ ],
1395
+ "widgets_values": [
1396
+ "01_Elena.png",
1397
+ "image"
1398
+ ],
1399
+ "title": "Picture 3"
1400
+ },
1401
+ {
1402
+ "properties": {
1403
+ "Node name for S&R": "LoadImage"
1404
+ },
1405
+ "id": 44,
1406
+ "type": "LoadImage",
1407
+ "pos": [
1408
+ 1710,
1409
+ 790
1410
+ ],
1411
+ "size": [
1412
+ 270,
1413
+ 360
1414
+ ],
1415
+ "flags": {},
1416
+ "order": 30,
1417
+ "mode": 4,
1418
+ "inputs": [],
1419
+ "outputs": [
1420
+ {
1421
+ "name": "IMAGE",
1422
+ "type": "IMAGE",
1423
+ "links": [
1424
+ 20
1425
+ ]
1426
+ },
1427
+ {
1428
+ "name": "MASK",
1429
+ "type": "MASK",
1430
+ "links": []
1431
+ }
1432
+ ],
1433
+ "widgets_values": [
1434
+ "01_Elena.png",
1435
+ "image"
1436
+ ],
1437
+ "title": "Picture 4"
1438
+ },
1439
+ {
1440
+ "properties": {
1441
+ "Node name for S&R": "LoadImage"
1442
+ },
1443
+ "id": 45,
1444
+ "type": "LoadImage",
1445
+ "pos": [
1446
+ 2000,
1447
+ 790
1448
+ ],
1449
+ "size": [
1450
+ 270,
1451
+ 360
1452
+ ],
1453
+ "flags": {},
1454
+ "order": 31,
1455
+ "mode": 4,
1456
+ "inputs": [],
1457
+ "outputs": [
1458
+ {
1459
+ "name": "IMAGE",
1460
+ "type": "IMAGE",
1461
+ "links": [
1462
+ 21
1463
+ ]
1464
+ },
1465
+ {
1466
+ "name": "MASK",
1467
+ "type": "MASK",
1468
+ "links": []
1469
+ }
1470
+ ],
1471
+ "widgets_values": [
1472
+ "01_Elena.png",
1473
+ "image"
1474
+ ],
1475
+ "title": "Picture 5"
1476
+ },
1477
+ {
1478
+ "properties": {
1479
+ "Node name for S&R": "LoadImage"
1480
+ },
1481
+ "id": 46,
1482
+ "type": "LoadImage",
1483
+ "pos": [
1484
+ 840,
1485
+ 1210
1486
+ ],
1487
+ "size": [
1488
+ 270,
1489
+ 360
1490
+ ],
1491
+ "flags": {},
1492
+ "order": 32,
1493
+ "mode": 4,
1494
+ "inputs": [],
1495
+ "outputs": [
1496
+ {
1497
+ "name": "IMAGE",
1498
+ "type": "IMAGE",
1499
+ "links": [
1500
+ 22
1501
+ ]
1502
+ },
1503
+ {
1504
+ "name": "MASK",
1505
+ "type": "MASK",
1506
+ "links": []
1507
+ }
1508
+ ],
1509
+ "widgets_values": [
1510
+ "01_Elena.png",
1511
+ "image"
1512
+ ],
1513
+ "title": "Picture 6"
1514
+ },
1515
+ {
1516
+ "properties": {
1517
+ "Node name for S&R": "LoadImage"
1518
+ },
1519
+ "id": 47,
1520
+ "type": "LoadImage",
1521
+ "pos": [
1522
+ 1130,
1523
+ 1210
1524
+ ],
1525
+ "size": [
1526
+ 270,
1527
+ 360
1528
+ ],
1529
+ "flags": {},
1530
+ "order": 33,
1531
+ "mode": 4,
1532
+ "inputs": [],
1533
+ "outputs": [
1534
+ {
1535
+ "name": "IMAGE",
1536
+ "type": "IMAGE",
1537
+ "links": [
1538
+ 23
1539
+ ]
1540
+ },
1541
+ {
1542
+ "name": "MASK",
1543
+ "type": "MASK",
1544
+ "links": []
1545
+ }
1546
+ ],
1547
+ "widgets_values": [
1548
+ "01_Elena.png",
1549
+ "image"
1550
+ ],
1551
+ "title": "Picture 7"
1552
+ },
1553
+ {
1554
+ "properties": {
1555
+ "Node name for S&R": "LoadImage"
1556
+ },
1557
+ "id": 48,
1558
+ "type": "LoadImage",
1559
+ "pos": [
1560
+ 1420,
1561
+ 1210
1562
+ ],
1563
+ "size": [
1564
+ 270,
1565
+ 360
1566
+ ],
1567
+ "flags": {},
1568
+ "order": 34,
1569
+ "mode": 4,
1570
+ "inputs": [],
1571
+ "outputs": [
1572
+ {
1573
+ "name": "IMAGE",
1574
+ "type": "IMAGE",
1575
+ "links": [
1576
+ 24
1577
+ ]
1578
+ },
1579
+ {
1580
+ "name": "MASK",
1581
+ "type": "MASK",
1582
+ "links": []
1583
+ }
1584
+ ],
1585
+ "widgets_values": [
1586
+ "01_Elena.png",
1587
+ "image"
1588
+ ],
1589
+ "title": "Picture 8"
1590
+ },
1591
+ {
1592
+ "properties": {
1593
+ "Node name for S&R": "LoadImage"
1594
+ },
1595
+ "id": 49,
1596
+ "type": "LoadImage",
1597
+ "pos": [
1598
+ 1710,
1599
+ 1210
1600
+ ],
1601
+ "size": [
1602
+ 270,
1603
+ 360
1604
+ ],
1605
+ "flags": {},
1606
+ "order": 35,
1607
+ "mode": 4,
1608
+ "inputs": [],
1609
+ "outputs": [
1610
+ {
1611
+ "name": "IMAGE",
1612
+ "type": "IMAGE",
1613
+ "links": [
1614
+ 25
1615
+ ]
1616
+ },
1617
+ {
1618
+ "name": "MASK",
1619
+ "type": "MASK",
1620
+ "links": []
1621
+ }
1622
+ ],
1623
+ "widgets_values": [
1624
+ "01_Elena.png",
1625
+ "image"
1626
+ ],
1627
+ "title": "Picture 9"
1628
+ },
1629
+ {
1630
+ "properties": {
1631
+ "Node name for S&R": "LoadImage"
1632
+ },
1633
+ "id": 50,
1634
+ "type": "LoadImage",
1635
+ "pos": [
1636
+ 2000,
1637
+ 1210
1638
+ ],
1639
+ "size": [
1640
+ 270,
1641
+ 360
1642
+ ],
1643
+ "flags": {},
1644
+ "order": 36,
1645
+ "mode": 4,
1646
+ "inputs": [],
1647
+ "outputs": [
1648
+ {
1649
+ "name": "IMAGE",
1650
+ "type": "IMAGE",
1651
+ "links": [
1652
+ 26
1653
+ ]
1654
+ },
1655
+ {
1656
+ "name": "MASK",
1657
+ "type": "MASK",
1658
+ "links": []
1659
+ }
1660
+ ],
1661
+ "widgets_values": [
1662
+ "01_Elena.png",
1663
+ "image"
1664
+ ],
1665
+ "title": "Picture 10"
1666
+ },
1667
+ {
1668
+ "properties": {
1669
+ "Node name for S&R": "MarkdownNote"
1670
+ },
1671
+ "id": 60,
1672
+ "type": "MarkdownNote",
1673
+ "pos": [
1674
+ -570,
1675
+ 0
1676
+ ],
1677
+ "size": [
1678
+ 520,
1679
+ 1100
1680
+ ],
1681
+ "flags": {},
1682
+ "order": 37,
1683
+ "mode": 0,
1684
+ "inputs": [],
1685
+ "outputs": [],
1686
+ "widgets_values": [
1687
+ "## MiniMax H3 Image - 16 MP on 8 GB VRAM\n\nSame model, sampler and **16 MP** final size as the 16 GB version. ComfyUI streams the model through the card in parts, so a smaller card gives the **same image**, only a little slower. The VAE decode is tiled to keep the last step small.\n\n**1. Prompt** - write it **long**, like a real photo: who (age, hair, skin, outfit), where, what is around them, the light and its direction, and the camera, e.g. *\"Shot on the iPhone 16 Pro front camera, visible skin texture and pores, natural phone HDR, no retouching, ordinary Instagram photo, not cinematic, not studio.\"* Short prompts give a glossy doll face. *Enhance prompt* (Qwen3.5 upsampler) can write the long version for text-to-image; leave it off with reference pictures.\n\n**2. Size** - pick the ASPECT RATIO. *16 MP* on = final, off = 4 MP draft for testing prompts fast. People: use 3:4 (or 4:3).\n\n| Ratio | draft 4 MP | 16 MP |\n|---|---|---|\n| 1:1 | 2048x2048 | 4096x4096 |\n| 3:4 | 1760x2368 | 3552x4736 |\n| 4:3 | 2368x1760 | 4736x3552 |\n| 2:3 | 1664x2496 | 3360x5024 |\n| 3:2 | 2496x1664 | 5024x3360 |\n| 9:16 | 1536x2720 | 3072x5472 |\n| 16:9 | 2720x1536 | 5472x3072 |\n| 21:9 | 3136x1344 | 6272x2688 |\n\n**3. Reference pictures** - 10 slots, all bypassed. Select a slot, press **Ctrl+B** to switch it on, and load your picture. Fill slots from 1 upward: with slots 1, 2 and 3 on they are `<Picture 1>`, `<Picture 2>`, `<Picture 3>` in the prompt. Picture 1 is the one an edit instruction changes. Keep the VAE connected to the conditioning node so the face and room do not drift.\n\n**4. Sampler** - SamplerLCM (s_noise 1.1 -> 0.9), BasicScheduler **simple, 6 steps**, CFGGuider **cfg 1.0**. Do not raise the steps: the model is distilled for 6.\n\n**Decode** - VAE Decode (tiled), tile 512. If the decode runs out of memory, lower tile_size to 384 or 256.\n\n**Measured on 8 GB** (RTX 5070 Ti limited to the VRAM an 8 GB card has free, same seed gives the identical image as on 16 GB):\n- Text-to-image at **16 MP: about 1 minute** (5.7 s per step).\n- With **4 reference pictures: about 4 minutes at 16 MP**, about 2.5 minutes at 8 MP. The pictures fill the card, so each step takes longer.\n- A real 8 GB card (RTX 4060, 3070 ...) has a slower chip than the test card, so expect longer times than these.\n\n**8 GB tips**\n- Use the **4 MP draft** while you work on the prompt, then switch *16 MP* on for the final image with the same seed.\n- All **10 reference slots** work, but every picture makes each step slower: **4 or fewer** is the sweet spot. With many pictures, use 8 MP (set *Full megapixels* to 8) to save time.\n- Close other apps that use the GPU (games, video in the browser, OBS preview).\n- System RAM: tested with 24 GB. The models (21 GB + 16 GB) are read from disk in parts, so a fast SSD helps.\n"
1688
+ ],
1689
+ "title": "How to use (8 GB VRAM)"
1690
+ }
1691
+ ],
1692
+ "links": [
1693
+ [
1694
+ 1,
1695
+ 14,
1696
+ 0,
1697
+ 15,
1698
+ 0,
1699
+ "STRING"
1700
+ ],
1701
+ [
1702
+ 2,
1703
+ 10,
1704
+ 0,
1705
+ 15,
1706
+ 2,
1707
+ "STRING"
1708
+ ],
1709
+ [
1710
+ 3,
1711
+ 12,
1712
+ 0,
1713
+ 16,
1714
+ 0,
1715
+ "CLIP"
1716
+ ],
1717
+ [
1718
+ 4,
1719
+ 13,
1720
+ 0,
1721
+ 16,
1722
+ 1,
1723
+ "STRING"
1724
+ ],
1725
+ [
1726
+ 5,
1727
+ 15,
1728
+ 0,
1729
+ 16,
1730
+ 2,
1731
+ "STRING"
1732
+ ],
1733
+ [
1734
+ 6,
1735
+ 16,
1736
+ 0,
1737
+ 17,
1738
+ 0,
1739
+ "STRING"
1740
+ ],
1741
+ [
1742
+ 7,
1743
+ 10,
1744
+ 0,
1745
+ 18,
1746
+ 0,
1747
+ "STRING"
1748
+ ],
1749
+ [
1750
+ 8,
1751
+ 17,
1752
+ 0,
1753
+ 18,
1754
+ 1,
1755
+ "STRING"
1756
+ ],
1757
+ [
1758
+ 9,
1759
+ 11,
1760
+ 0,
1761
+ 18,
1762
+ 2,
1763
+ "BOOLEAN"
1764
+ ],
1765
+ [
1766
+ 10,
1767
+ 18,
1768
+ 0,
1769
+ 19,
1770
+ 0,
1771
+ "STRING"
1772
+ ],
1773
+ [
1774
+ 11,
1775
+ 21,
1776
+ 0,
1777
+ 23,
1778
+ 0,
1779
+ "FLOAT"
1780
+ ],
1781
+ [
1782
+ 12,
1783
+ 22,
1784
+ 0,
1785
+ 23,
1786
+ 1,
1787
+ "FLOAT"
1788
+ ],
1789
+ [
1790
+ 13,
1791
+ 20,
1792
+ 0,
1793
+ 23,
1794
+ 2,
1795
+ "BOOLEAN"
1796
+ ],
1797
+ [
1798
+ 14,
1799
+ 23,
1800
+ 0,
1801
+ 24,
1802
+ 1,
1803
+ "FLOAT"
1804
+ ],
1805
+ [
1806
+ 15,
1807
+ 2,
1808
+ 0,
1809
+ 30,
1810
+ 0,
1811
+ "CLIP"
1812
+ ],
1813
+ [
1814
+ 16,
1815
+ 3,
1816
+ 0,
1817
+ 30,
1818
+ 1,
1819
+ "VAE"
1820
+ ],
1821
+ [
1822
+ 17,
1823
+ 41,
1824
+ 0,
1825
+ 30,
1826
+ 2,
1827
+ "IMAGE"
1828
+ ],
1829
+ [
1830
+ 18,
1831
+ 42,
1832
+ 0,
1833
+ 30,
1834
+ 3,
1835
+ "IMAGE"
1836
+ ],
1837
+ [
1838
+ 19,
1839
+ 43,
1840
+ 0,
1841
+ 30,
1842
+ 4,
1843
+ "IMAGE"
1844
+ ],
1845
+ [
1846
+ 20,
1847
+ 44,
1848
+ 0,
1849
+ 30,
1850
+ 5,
1851
+ "IMAGE"
1852
+ ],
1853
+ [
1854
+ 21,
1855
+ 45,
1856
+ 0,
1857
+ 30,
1858
+ 6,
1859
+ "IMAGE"
1860
+ ],
1861
+ [
1862
+ 22,
1863
+ 46,
1864
+ 0,
1865
+ 30,
1866
+ 7,
1867
+ "IMAGE"
1868
+ ],
1869
+ [
1870
+ 23,
1871
+ 47,
1872
+ 0,
1873
+ 30,
1874
+ 8,
1875
+ "IMAGE"
1876
+ ],
1877
+ [
1878
+ 24,
1879
+ 48,
1880
+ 0,
1881
+ 30,
1882
+ 9,
1883
+ "IMAGE"
1884
+ ],
1885
+ [
1886
+ 25,
1887
+ 49,
1888
+ 0,
1889
+ 30,
1890
+ 10,
1891
+ "IMAGE"
1892
+ ],
1893
+ [
1894
+ 26,
1895
+ 50,
1896
+ 0,
1897
+ 30,
1898
+ 11,
1899
+ "IMAGE"
1900
+ ],
1901
+ [
1902
+ 27,
1903
+ 18,
1904
+ 0,
1905
+ 30,
1906
+ 12,
1907
+ "STRING"
1908
+ ],
1909
+ [
1910
+ 28,
1911
+ 24,
1912
+ 0,
1913
+ 30,
1914
+ 13,
1915
+ "INT"
1916
+ ],
1917
+ [
1918
+ 29,
1919
+ 24,
1920
+ 1,
1921
+ 30,
1922
+ 14,
1923
+ "INT"
1924
+ ],
1925
+ [
1926
+ 30,
1927
+ 30,
1928
+ 0,
1929
+ 31,
1930
+ 0,
1931
+ "CONDITIONING"
1932
+ ],
1933
+ [
1934
+ 31,
1935
+ 3,
1936
+ 0,
1937
+ 33,
1938
+ 1,
1939
+ "VAE"
1940
+ ],
1941
+ [
1942
+ 32,
1943
+ 33,
1944
+ 0,
1945
+ 34,
1946
+ 0,
1947
+ "IMAGE"
1948
+ ],
1949
+ [
1950
+ 33,
1951
+ 1,
1952
+ 0,
1953
+ 36,
1954
+ 0,
1955
+ "MODEL"
1956
+ ],
1957
+ [
1958
+ 34,
1959
+ 30,
1960
+ 0,
1961
+ 36,
1962
+ 1,
1963
+ "CONDITIONING"
1964
+ ],
1965
+ [
1966
+ 35,
1967
+ 31,
1968
+ 0,
1969
+ 36,
1970
+ 2,
1971
+ "CONDITIONING"
1972
+ ],
1973
+ [
1974
+ 36,
1975
+ 1,
1976
+ 0,
1977
+ 37,
1978
+ 0,
1979
+ "MODEL"
1980
+ ],
1981
+ [
1982
+ 37,
1983
+ 35,
1984
+ 0,
1985
+ 39,
1986
+ 0,
1987
+ "NOISE"
1988
+ ],
1989
+ [
1990
+ 38,
1991
+ 36,
1992
+ 0,
1993
+ 39,
1994
+ 1,
1995
+ "GUIDER"
1996
+ ],
1997
+ [
1998
+ 39,
1999
+ 38,
2000
+ 0,
2001
+ 39,
2002
+ 2,
2003
+ "SAMPLER"
2004
+ ],
2005
+ [
2006
+ 40,
2007
+ 37,
2008
+ 0,
2009
+ 39,
2010
+ 3,
2011
+ "SIGMAS"
2012
+ ],
2013
+ [
2014
+ 41,
2015
+ 30,
2016
+ 1,
2017
+ 39,
2018
+ 4,
2019
+ "LATENT"
2020
+ ],
2021
+ [
2022
+ 42,
2023
+ 39,
2024
+ 0,
2025
+ 33,
2026
+ 0,
2027
+ "LATENT"
2028
+ ]
2029
+ ],
2030
+ "groups": [
2031
+ {
2032
+ "id": 1,
2033
+ "title": "Models",
2034
+ "bounding": [
2035
+ -20,
2036
+ -80,
2037
+ 460,
2038
+ 480
2039
+ ],
2040
+ "color": "#3f789e",
2041
+ "font_size": 24,
2042
+ "flags": {}
2043
+ },
2044
+ {
2045
+ "id": 2,
2046
+ "title": "Prompt",
2047
+ "bounding": [
2048
+ -20,
2049
+ 440,
2050
+ 810,
2051
+ 1240
2052
+ ],
2053
+ "color": "#a1309b",
2054
+ "font_size": 24,
2055
+ "flags": {}
2056
+ },
2057
+ {
2058
+ "id": 3,
2059
+ "title": "Size: aspect ratio + 16 MP (8 GB VRAM)",
2060
+ "bounding": [
2061
+ 820,
2062
+ -80,
2063
+ 690,
2064
+ 520
2065
+ ],
2066
+ "color": "#88A",
2067
+ "font_size": 24,
2068
+ "flags": {}
2069
+ },
2070
+ {
2071
+ "id": 4,
2072
+ "title": "Reference pictures - 4 or fewer is fastest (Ctrl+B to switch a slot on)",
2073
+ "bounding": [
2074
+ 820,
2075
+ 710,
2076
+ 1470,
2077
+ 870
2078
+ ],
2079
+ "color": "#8A8",
2080
+ "font_size": 24,
2081
+ "flags": {}
2082
+ },
2083
+ {
2084
+ "id": 5,
2085
+ "title": "Render",
2086
+ "bounding": [
2087
+ 1540,
2088
+ -80,
2089
+ 1330,
2090
+ 750
2091
+ ],
2092
+ "color": "#b58b2a",
2093
+ "font_size": 24,
2094
+ "flags": {}
2095
+ }
2096
+ ],
2097
+ "config": {},
2098
+ "extra": {
2099
+ "ds": {
2100
+ "scale": 0.5,
2101
+ "offset": [
2102
+ 600,
2103
+ 150
2104
+ ]
2105
+ }
2106
+ },
2107
+ "version": 0.4
2108
+ }