greenslime drbaph commited on
Commit
e3140cb
·
0 Parent(s):

Duplicate from drbaph/MiniMax-H3-Turbo-Lora-ComfyUI

Browse files

Co-authored-by: DRBAPH <drbaph@users.noreply.huggingface.co>

.gitattributes ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ assets/1.webp filter=lfs diff=lfs merge=lfs -text
37
+ assets/2.webp filter=lfs diff=lfs merge=lfs -text
38
+ assets/MiniMax_H3_00060-audio.mp4 filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,285 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: Comfy-Org/MiniMax-H3
4
+ base_model_relation: adapter
5
+ pipeline_tag: text-to-video
6
+ tags:
7
+ - minimax-h3
8
+ - lora
9
+ - adapter
10
+ - comfyui
11
+ - pruned
12
+ - pruned-model
13
+ - curve-form
14
+ - text-to-video
15
+ - text-to-audio
16
+ - audio-video
17
+ - video-generation
18
+ - audio-generation
19
+ - synchronized-audio
20
+ - four-step
21
+ - 4-step
22
+ - turbo
23
+ - accelerated-inference
24
+ - further-trained
25
+ - checkpoint-500
26
+ - safetensors
27
+ - bfloat16
28
+ - bf16
29
+ - partial-conversion
30
+ ---
31
+
32
+ # MiniMax-H3 Turbo 4-Step — ComfyUI Pruned-Model LoRAs
33
+
34
+ This repository contains third-party ComfyUI compatibility conversions of the original **MiniMax-H3 Turbo LoRA — 4-step audio-video generation preview** created by **[larryvrh](https://huggingface.co/larryvrh)**.
35
+
36
+ It also includes two **further-trained checkpoint-500 variants** in the same pruned-model ComfyUI format.
37
+
38
+ These LoRAs are intended for the **pruned/curve-form MiniMax-H3 checkpoint used by ComfyUI**.
39
+
40
+ **also check https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo**
41
+
42
+ ## ComfyUI Test Output
43
+
44
+ <video
45
+ src="https://huggingface.co/drbaph/MiniMax-H3-Turbo-Lora-ComfyUI/resolve/main/assets/MiniMax_H3_00060-audio.mp4"
46
+ controls
47
+ loop
48
+ width="100%">
49
+ </video>
50
+
51
+ [Download the ComfyUI workflow](https://huggingface.co/drbaph/MiniMax-H3-Turbo-Lora-ComfyUI/resolve/main/fl_minimax_h3_turbo_lora_example_workflow.json?download=true)
52
+
53
+ ![ComfyUI workflow screenshot](https://cdn-uploads.huggingface.co/production/uploads/63473b59e5c0717e6737b872/zfVV_yKP-njnOrGinGy8K.png)
54
+
55
+ ## Original Project
56
+
57
+ The original LoRA weights, training work, four-step distillation method, dual video/audio sampling implementation, documentation and source files were created and released by **larryvrh**:
58
+
59
+ **Original repository:**
60
+ [larryvrh/MiniMax-H3-Turbo-Lora](https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora)
61
+
62
+ I did **not** train, distill or create the original Turbo LoRA weights. This repository provides modified compatibility versions intended to allow the compatible LoRA adapters to load with ComfyUI's built-in MiniMax-H3 LoRA loader when using the pruned/curve-form model.
63
+
64
+ Full credit for the original MiniMax-H3 Turbo LoRA and its distillation method belongs to **larryvrh**.
65
+
66
+ ## Included Files
67
+
68
+ | File | Description |
69
+ |---|---|
70
+ | `minimax_h3_turbo_4step_pruned_comfyui.safetensors` | Initial partial ComfyUI compatibility conversion of the original non-EMA Turbo weights for the pruned/curve-form MiniMax-H3 checkpoint. The original author describes the non-EMA weights as crisper and better at holding fast motion. |
71
+ | `minimax_h3_turbo_4step_ema_pruned_comfyui.safetensors` | Initial partial ComfyUI compatibility conversion of the original EMA Turbo weights for the pruned/curve-form MiniMax-H3 checkpoint. The original author describes the early EMA weights as smoother but softer because the EMA had not fully matured. |
72
+ | `minimax_h3_turbo_4step_ckpt500_pruned_comfyui.safetensors` | Further-trained checkpoint-500 non-EMA variant in the pruned-model ComfyUI format. This continues training beyond the initial preview checkpoint. |
73
+ | `minimax_h3_turbo_4step_ema_ckpt500_pruned_comfyui.safetensors` | Further-trained checkpoint-500 EMA variant in the pruned-model ComfyUI format. This contains the time-averaged weights from the further-trained checkpoint. |
74
+ | `fl_minimax_h3_turbo_lora_example_workflow.json` | First-and-last-frame ComfyUI example workflow using the Turbo LoRA, with optional attention and memory optimizations. |
75
+
76
+ The EMA and non-EMA files remain separate and contain different tensor values.
77
+
78
+ The original converted files preserve the initial preview weights, while the `ckpt500` files provide the further-trained variants.
79
+
80
+ These are partial compatibility versions of the original full-model Turbo LoRAs for use with the pruned/curve-form MiniMax-H3 architecture.
81
+
82
+ ## Which File Should I Use?
83
+
84
+ For the further-trained weights, start with:
85
+
86
+ `minimax_h3_turbo_4step_ckpt500_pruned_comfyui.safetensors`
87
+
88
+ Use the further-trained EMA variant for comparison:
89
+
90
+ `minimax_h3_turbo_4step_ema_ckpt500_pruned_comfyui.safetensors`
91
+
92
+ The two files without `ckpt500` preserve the earlier preview versions:
93
+
94
+ - `minimax_h3_turbo_4step_pruned_comfyui.safetensors`
95
+ - `minimax_h3_turbo_4step_ema_pruned_comfyui.safetensors`
96
+
97
+ The non-EMA variants generally preserve a sharper response and stronger fast-motion behavior, while the EMA variants are intended to provide smoother time-averaged weights.
98
+
99
+ ## About the Original Turbo LoRA
100
+
101
+ The original MiniMax-H3 Turbo LoRA is an early preview checkpoint designed to generate joint video and synchronized stereo audio in approximately **4 sampling steps instead of the usual approximately 20 steps**, providing roughly a **5× reduction in sampling wall-clock time**.
102
+
103
+ The original author clearly states that the initial release is an unfinished preview checkpoint from an in-progress training run:
104
+
105
+ - it is under-trained;
106
+ - the initial EMA weights have not fully matured;
107
+ - quality is not representative of a completed run;
108
+ - it is intended as an early demonstration of the direction of the project.
109
+
110
+ The additional `ckpt500` files in this repository are further-trained variants beyond that initial preview checkpoint.
111
+
112
+ Refer to the original repository for the original author's documentation, generation script, scheduler implementation and future updates:
113
+
114
+ [larryvrh/MiniMax-H3-Turbo-Lora](https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora)
115
+
116
+ ## Why This Conversion Was Made
117
+
118
+ The original Turbo LoRA files use tensor keys such as:
119
+
120
+ `blocks.0.attn.qkv_proj.lora_A.weight`
121
+
122
+ ComfyUI's built-in MiniMax-H3 LoRA resolver expects keys under the model namespace:
123
+
124
+ `diffusion_model.blocks.0.attn.qkv_proj.lora_A.weight`
125
+
126
+ Without that namespace, ComfyUI reports warnings such as:
127
+
128
+ `[WARNING] lora key not loaded: blocks.0.attn.out_proj.lora_A.weight`
129
+
130
+ The following namespace conversion was applied to the retained tensors:
131
+
132
+ - `blocks.*` → `diffusion_model.blocks.*`
133
+ - `token_refiner.*` → `diffusion_model.token_refiner.*`
134
+
135
+ The retained tensor values, BF16 dtype, tensor shapes, LoRA rank and A/B orientation were preserved.
136
+
137
+ ## Conversion Details
138
+
139
+ Each original full-model Turbo LoRA contained:
140
+
141
+ | Component | Count |
142
+ |---|---:|
143
+ | Source tensors | 518 |
144
+ | Source LoRA A/B pairs | 259 |
145
+
146
+ Each pruned-model compatibility conversion contains:
147
+
148
+ | Component | Count |
149
+ |---|---:|
150
+ | Retained tensors | 416 |
151
+ | Retained LoRA A/B pairs | 208 |
152
+ | Pruned AdaLN tensors | 102 |
153
+ | Pruned AdaLN A/B pairs | 51 |
154
+
155
+ The pruned adapters were:
156
+
157
+ - 50 AdaLN projection pairs from `blocks.0` through `blocks.49`;
158
+ - 1 AdaLN projection pair from `final_layer.adaln_proj.linear`.
159
+
160
+ The retained adapters include:
161
+
162
+ - all main-block attention adapters;
163
+ - all main-block MLP adapters;
164
+ - all token-refiner attention adapters;
165
+ - all token-refiner MLP adapters.
166
+
167
+ The retained attention, MLP and token-refiner tensors have dimensions compatible with the pruned/curve-form MiniMax-H3 checkpoint.
168
+
169
+ ## Validation
170
+
171
+ The converted LoRA structure was checked for:
172
+
173
+ - complete LoRA A/B adapter pairs;
174
+ - no orphan A or B tensors;
175
+ - retained keys using the `diffusion_model.*` namespace;
176
+ - no duplicated `diffusion_model.diffusion_model.*` prefixes;
177
+ - no incompatible `adaln_proj` tensors;
178
+ - preserved BF16 dtype;
179
+ - preserved tensor shapes;
180
+ - preserved LoRA ranks;
181
+ - separate EMA and non-EMA weights;
182
+ - no unresolved retained keys during ComfyUI loader-level validation.
183
+
184
+ The original source files were not overwritten.
185
+
186
+ The LoRAs have been tested working in ComfyUI with the following optional acceleration methods:
187
+
188
+ - **SageAttention**
189
+ - **Sol Attention**
190
+ - **Gradient**
191
+ - **Spectrum**
192
+
193
+ These acceleration methods are optional and are not included in this repository.
194
+
195
+ ## Important Limitation
196
+
197
+ These files are **partial ComfyUI compatibility versions for the pruned/curve-form MiniMax-H3 checkpoint**.
198
+
199
+ The further-trained `ckpt500` variants improve upon the initial compatibility versions,
200
+
201
+ ## Usage
202
+
203
+ Place the LoRA files inside:
204
+
205
+ `ComfyUI/models/loras/`
206
+
207
+ Use them with the **pruned/curve-form MiniMax-H3 checkpoint for ComfyUI**.
208
+
209
+ Recommended further-trained version:
210
+
211
+ `minimax_h3_turbo_4step_ckpt500_pruned_comfyui.safetensors`
212
+
213
+ Further-trained EMA version:
214
+
215
+ `minimax_h3_turbo_4step_ema_ckpt500_pruned_comfyui.safetensors`
216
+
217
+ Suggested ComfyUI settings:
218
+
219
+ - Video sigma shift: `12`
220
+ - Audio sigma shift: `4-6`
221
+ - Steps tested working: `8–10` / `6-8` (ckpt500)
222
+ - Recommended sampler: `res_multistep`
223
+ - Suggested LoRA strength: `0.8–1.8`
224
+ - A higher LoRA strength can generally be used with fewer sampling steps
225
+ - Use the same prompt, seed, resolution, model checkpoint, sampler and workflow when comparing files
226
+
227
+ Optional accelerators confirmed working:
228
+
229
+ - **SageAttention**
230
+ - **Sol Attention**
231
+ - **Gradient**
232
+ - **Spectrum**
233
+
234
+ Do not assume broken audio is caused by the LoRA before checking the sampler, scheduler and audio sigma-shift configuration.
235
+
236
+ ## Audio and Scheduler Notes
237
+
238
+ MiniMax-H3 generates video and synchronized audio as separate streams with different scheduling requirements.
239
+
240
+ The original author's standalone generation implementation uses a dual video/audio scheduling system specifically designed for ultra-low-step generation.
241
+
242
+ For this ComfyUI pruned-model version, the tested settings are:
243
+
244
+ - Video sigma shift: `12`
245
+ - Audio sigma shift: `4-6`
246
+ - Steps: `8–10` / `6-8` (ckpt500)
247
+ - Sampler: `res_multistep`
248
+
249
+ Incorrect audio scheduling may result in:
250
+
251
+ - blown-out audio;
252
+ - severe distortion;
253
+ - noise-like audio;
254
+ - unstable audio energy;
255
+ - completely broken audio.
256
+
257
+ For the original full-model settings, model files, resolution guidance, frame-count rules and dual-scheduler implementation, follow the original README:
258
+
259
+ [larryvrh/MiniMax-H3-Turbo-Lora](https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora)
260
+
261
+ ## Attribution
262
+
263
+ Original MiniMax-H3 Turbo LoRA author:
264
+
265
+ **[larryvrh](https://huggingface.co/larryvrh)**
266
+
267
+ Original project:
268
+
269
+ **[MiniMax-H3-Turbo-Lora](https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora)**
270
+
271
+ This repository contains third-party, partial ComfyUI compatibility versions for the pruned/curve-form MiniMax-H3 checkpoint, including further-trained checkpoint-500 variants.
272
+
273
+ I do not claim authorship of:
274
+
275
+ - the original Turbo LoRA weights;
276
+ - the original LoRA training;
277
+ - the original distillation method;
278
+ - the original four-step sampling method;
279
+ - the original dual video/audio scheduler;
280
+ - the original `generate.py`;
281
+ - the original documentation;
282
+ - the MiniMax-H3 base model;
283
+ - the pruned/curve-form MiniMax-H3 checkpoint.
284
+
285
+ Please retain attribution to **larryvrh** and comply with the licences and terms of the original LoRA and MiniMax-H3 base model when using or redistributing these files.
assets/1.webp ADDED

Git LFS Details

  • SHA256: 669addd55c4fb48afc4ae21f38d0f3b2744679275473bade6ec8297f04f86104
  • Pointer size: 131 Bytes
  • Size of remote file: 250 kB
assets/2.webp ADDED

Git LFS Details

  • SHA256: 84bdbcc616cc35ef3cf3e0080c4f366d24d5002f9fa897eb87fd163613ad9ef1
  • Pointer size: 131 Bytes
  • Size of remote file: 245 kB
assets/MiniMax_H3_00060-audio.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63a036f90f5aa07b65e64bf3a6cba8ba1ca02d3923c1bc86e2f127d37d309d63
3
+ size 12557337
fl_minimax_h3_turbo_lora_example_workflow.json ADDED
@@ -0,0 +1,1815 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "id": "e3f2b845-8f2c-4b5a-9caf-eac1029d3e7e",
3
+ "revision": 0,
4
+ "last_node_id": 146,
5
+ "last_link_id": 269,
6
+ "nodes": [
7
+ {
8
+ "id": 125,
9
+ "type": "KSamplerSelect",
10
+ "pos": [
11
+ -335.1773390857757,
12
+ 4897.280631458884
13
+ ],
14
+ "size": [
15
+ 370,
16
+ 70
17
+ ],
18
+ "flags": {},
19
+ "order": 0,
20
+ "mode": 0,
21
+ "inputs": [],
22
+ "outputs": [
23
+ {
24
+ "name": "SAMPLER",
25
+ "type": "SAMPLER",
26
+ "links": [
27
+ 234
28
+ ]
29
+ }
30
+ ],
31
+ "properties": {
32
+ "cnr_id": "comfy-core",
33
+ "ver": "0.30.0",
34
+ "Node name for S&R": "KSamplerSelect",
35
+ "ue_properties": {
36
+ "widget_ue_connectable": {},
37
+ "version": "7.8",
38
+ "input_ue_unconnectable": {}
39
+ }
40
+ },
41
+ "widgets_values": [
42
+ "res_multistep"
43
+ ]
44
+ },
45
+ {
46
+ "id": 127,
47
+ "type": "SamplerCustomAdvanced",
48
+ "pos": [
49
+ 84.82266091422431,
50
+ 4807.280631458884
51
+ ],
52
+ "size": [
53
+ 230,
54
+ 140
55
+ ],
56
+ "flags": {},
57
+ "order": 26,
58
+ "mode": 0,
59
+ "inputs": [
60
+ {
61
+ "name": "noise",
62
+ "type": "NOISE",
63
+ "link": 232
64
+ },
65
+ {
66
+ "name": "guider",
67
+ "type": "GUIDER",
68
+ "link": 233
69
+ },
70
+ {
71
+ "name": "sampler",
72
+ "type": "SAMPLER",
73
+ "link": 234
74
+ },
75
+ {
76
+ "name": "sigmas",
77
+ "type": "SIGMAS",
78
+ "link": 235
79
+ },
80
+ {
81
+ "name": "latent_image",
82
+ "type": "LATENT",
83
+ "link": 236
84
+ }
85
+ ],
86
+ "outputs": [
87
+ {
88
+ "name": "output",
89
+ "type": "LATENT",
90
+ "links": [
91
+ 249,
92
+ 250
93
+ ]
94
+ },
95
+ {
96
+ "name": "denoised_output",
97
+ "type": "LATENT",
98
+ "links": []
99
+ }
100
+ ],
101
+ "properties": {
102
+ "cnr_id": "comfy-core",
103
+ "ver": "0.30.0",
104
+ "Node name for S&R": "SamplerCustomAdvanced",
105
+ "ue_properties": {
106
+ "widget_ue_connectable": {},
107
+ "version": "7.8",
108
+ "input_ue_unconnectable": {}
109
+ }
110
+ },
111
+ "widgets_values": []
112
+ },
113
+ {
114
+ "id": 130,
115
+ "type": "CLIPLoader",
116
+ "pos": [
117
+ -1570.615933590389,
118
+ 4836.578902898876
119
+ ],
120
+ "size": [
121
+ 640,
122
+ 120
123
+ ],
124
+ "flags": {},
125
+ "order": 1,
126
+ "mode": 0,
127
+ "inputs": [],
128
+ "outputs": [
129
+ {
130
+ "name": "CLIP",
131
+ "type": "CLIP",
132
+ "links": [
133
+ 241
134
+ ]
135
+ }
136
+ ],
137
+ "properties": {
138
+ "cnr_id": "comfy-core",
139
+ "ver": "0.30.0",
140
+ "Node name for S&R": "CLIPLoader",
141
+ "ue_properties": {
142
+ "widget_ue_connectable": {},
143
+ "version": "7.8",
144
+ "input_ue_unconnectable": {}
145
+ },
146
+ "models": [
147
+ {
148
+ "name": "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
149
+ "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
150
+ "directory": "text_encoders"
151
+ }
152
+ ]
153
+ },
154
+ "widgets_values": [
155
+ "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
156
+ "minimax",
157
+ "default"
158
+ ]
159
+ },
160
+ {
161
+ "id": 124,
162
+ "type": "VAEDecode",
163
+ "pos": [
164
+ 414.8226609142243,
165
+ 4797.280631458884
166
+ ],
167
+ "size": [
168
+ 230,
169
+ 60
170
+ ],
171
+ "flags": {
172
+ "collapsed": false
173
+ },
174
+ "order": 27,
175
+ "mode": 0,
176
+ "inputs": [
177
+ {
178
+ "name": "samples",
179
+ "type": "LATENT",
180
+ "link": 249
181
+ },
182
+ {
183
+ "name": "vae",
184
+ "type": "VAE",
185
+ "link": 230
186
+ }
187
+ ],
188
+ "outputs": [
189
+ {
190
+ "name": "IMAGE",
191
+ "type": "IMAGE",
192
+ "links": [
193
+ 255
194
+ ]
195
+ }
196
+ ],
197
+ "properties": {
198
+ "cnr_id": "comfy-core",
199
+ "ver": "0.30.0",
200
+ "Node name for S&R": "VAEDecode",
201
+ "ue_properties": {
202
+ "widget_ue_connectable": {},
203
+ "version": "7.8",
204
+ "input_ue_unconnectable": {}
205
+ }
206
+ },
207
+ "widgets_values": []
208
+ },
209
+ {
210
+ "id": 123,
211
+ "type": "VAEDecodeAudio",
212
+ "pos": [
213
+ 414.8226609142243,
214
+ 4917.280631458884
215
+ ],
216
+ "size": [
217
+ 230,
218
+ 60
219
+ ],
220
+ "flags": {
221
+ "collapsed": false
222
+ },
223
+ "order": 28,
224
+ "mode": 0,
225
+ "inputs": [
226
+ {
227
+ "name": "samples",
228
+ "type": "LATENT",
229
+ "link": 250
230
+ },
231
+ {
232
+ "name": "vae",
233
+ "type": "VAE",
234
+ "link": 229
235
+ }
236
+ ],
237
+ "outputs": [
238
+ {
239
+ "name": "AUDIO",
240
+ "type": "AUDIO",
241
+ "links": [
242
+ 256
243
+ ]
244
+ }
245
+ ],
246
+ "properties": {
247
+ "cnr_id": "comfy-core",
248
+ "ver": "0.30.0",
249
+ "Node name for S&R": "VAEDecodeAudio",
250
+ "ue_properties": {
251
+ "widget_ue_connectable": {},
252
+ "version": "7.8",
253
+ "input_ue_unconnectable": {}
254
+ }
255
+ },
256
+ "widgets_values": []
257
+ },
258
+ {
259
+ "id": 128,
260
+ "type": "BasicGuider",
261
+ "pos": [
262
+ -335.1773390857757,
263
+ 4767.280631458884
264
+ ],
265
+ "size": [
266
+ 360,
267
+ 60
268
+ ],
269
+ "flags": {},
270
+ "order": 24,
271
+ "mode": 0,
272
+ "inputs": [
273
+ {
274
+ "name": "model",
275
+ "type": "MODEL",
276
+ "link": 264
277
+ },
278
+ {
279
+ "name": "conditioning",
280
+ "type": "CONDITIONING",
281
+ "link": 238
282
+ }
283
+ ],
284
+ "outputs": [
285
+ {
286
+ "name": "GUIDER",
287
+ "type": "GUIDER",
288
+ "links": [
289
+ 233
290
+ ]
291
+ }
292
+ ],
293
+ "properties": {
294
+ "cnr_id": "comfy-core",
295
+ "ver": "0.30.0",
296
+ "Node name for S&R": "BasicGuider",
297
+ "ue_properties": {
298
+ "widget_ue_connectable": {},
299
+ "version": "7.8",
300
+ "input_ue_unconnectable": {}
301
+ }
302
+ },
303
+ "widgets_values": []
304
+ },
305
+ {
306
+ "id": 131,
307
+ "type": "RandomNoise",
308
+ "pos": [
309
+ -335.1773390857757,
310
+ 4607.280631458884
311
+ ],
312
+ "size": [
313
+ 360,
314
+ 90
315
+ ],
316
+ "flags": {},
317
+ "order": 2,
318
+ "mode": 0,
319
+ "inputs": [],
320
+ "outputs": [
321
+ {
322
+ "name": "NOISE",
323
+ "type": "NOISE",
324
+ "links": [
325
+ 232
326
+ ]
327
+ }
328
+ ],
329
+ "properties": {
330
+ "cnr_id": "comfy-core",
331
+ "ver": "0.30.0",
332
+ "Node name for S&R": "RandomNoise",
333
+ "ue_properties": {
334
+ "widget_ue_connectable": {},
335
+ "version": "7.8",
336
+ "input_ue_unconnectable": {}
337
+ }
338
+ },
339
+ "widgets_values": [
340
+ 9583685594385,
341
+ "randomize"
342
+ ]
343
+ },
344
+ {
345
+ "id": 126,
346
+ "type": "BasicScheduler",
347
+ "pos": [
348
+ -335.1773390857757,
349
+ 5047.280631458884
350
+ ],
351
+ "size": [
352
+ 370,
353
+ 130
354
+ ],
355
+ "flags": {},
356
+ "order": 25,
357
+ "mode": 0,
358
+ "inputs": [
359
+ {
360
+ "name": "model",
361
+ "type": "MODEL",
362
+ "link": 265
363
+ }
364
+ ],
365
+ "outputs": [
366
+ {
367
+ "name": "SIGMAS",
368
+ "type": "SIGMAS",
369
+ "links": [
370
+ 235
371
+ ]
372
+ }
373
+ ],
374
+ "properties": {
375
+ "cnr_id": "comfy-core",
376
+ "ver": "0.30.0",
377
+ "Node name for S&R": "BasicScheduler",
378
+ "ue_properties": {
379
+ "widget_ue_connectable": {},
380
+ "version": "7.8",
381
+ "input_ue_unconnectable": {}
382
+ }
383
+ },
384
+ "widgets_values": [
385
+ "simple",
386
+ 10,
387
+ 1
388
+ ]
389
+ },
390
+ {
391
+ "id": 134,
392
+ "type": "ComfyMathExpression",
393
+ "pos": [
394
+ -1120.0600911450272,
395
+ 5318.974494680876
396
+ ],
397
+ "size": [
398
+ 360,
399
+ 160
400
+ ],
401
+ "flags": {
402
+ "collapsed": true
403
+ },
404
+ "order": 13,
405
+ "mode": 0,
406
+ "inputs": [
407
+ {
408
+ "label": "a",
409
+ "name": "values.a",
410
+ "type": "FLOAT,INT,BOOLEAN",
411
+ "link": 248
412
+ },
413
+ {
414
+ "label": "b",
415
+ "name": "values.b",
416
+ "shape": 7,
417
+ "type": "FLOAT,INT,BOOLEAN",
418
+ "link": null
419
+ }
420
+ ],
421
+ "outputs": [
422
+ {
423
+ "name": "FLOAT",
424
+ "type": "FLOAT",
425
+ "links": []
426
+ },
427
+ {
428
+ "name": "INT",
429
+ "type": "INT",
430
+ "links": [
431
+ 245
432
+ ]
433
+ },
434
+ {
435
+ "name": "BOOL",
436
+ "type": "BOOLEAN",
437
+ "links": []
438
+ }
439
+ ],
440
+ "properties": {
441
+ "cnr_id": "comfy-core",
442
+ "ver": "0.30.0",
443
+ "Node name for S&R": "ComfyMathExpression",
444
+ "ue_properties": {
445
+ "widget_ue_connectable": {},
446
+ "version": "7.8",
447
+ "input_ue_unconnectable": {}
448
+ }
449
+ },
450
+ "widgets_values": [
451
+ "max(5, round(a * 24)) + (5 - (max(5, round(a * 24)) % 17)) % 17"
452
+ ]
453
+ },
454
+ {
455
+ "id": 133,
456
+ "type": "MiniMaxH3ImageToVideo",
457
+ "pos": [
458
+ -845.177339085776,
459
+ 4577.280631458884
460
+ ],
461
+ "size": [
462
+ 430,
463
+ 550
464
+ ],
465
+ "flags": {},
466
+ "order": 18,
467
+ "mode": 0,
468
+ "inputs": [
469
+ {
470
+ "name": "clip",
471
+ "type": "CLIP",
472
+ "link": 241
473
+ },
474
+ {
475
+ "name": "vae",
476
+ "type": "VAE",
477
+ "link": 242
478
+ },
479
+ {
480
+ "name": "first_frame",
481
+ "shape": 7,
482
+ "type": "IMAGE",
483
+ "link": 261
484
+ },
485
+ {
486
+ "name": "last_frame",
487
+ "shape": 7,
488
+ "type": "IMAGE",
489
+ "link": 269
490
+ },
491
+ {
492
+ "name": "width",
493
+ "type": "INT",
494
+ "widget": {
495
+ "name": "width"
496
+ },
497
+ "link": 246
498
+ },
499
+ {
500
+ "name": "height",
501
+ "type": "INT",
502
+ "widget": {
503
+ "name": "height"
504
+ },
505
+ "link": 247
506
+ },
507
+ {
508
+ "name": "length",
509
+ "type": "INT",
510
+ "widget": {
511
+ "name": "length"
512
+ },
513
+ "link": 245
514
+ }
515
+ ],
516
+ "outputs": [
517
+ {
518
+ "name": "positive",
519
+ "type": "CONDITIONING",
520
+ "links": [
521
+ 238
522
+ ]
523
+ },
524
+ {
525
+ "name": "LATENT",
526
+ "type": "LATENT",
527
+ "links": [
528
+ 236
529
+ ]
530
+ }
531
+ ],
532
+ "properties": {
533
+ "cnr_id": "comfy-core",
534
+ "ver": "0.30.0",
535
+ "Node name for S&R": "MiniMaxH3ImageToVideo",
536
+ "ue_properties": {
537
+ "widget_ue_connectable": {},
538
+ "version": "7.8",
539
+ "input_ue_unconnectable": {}
540
+ }
541
+ },
542
+ "widgets_values": [
543
+ "subject_definitions:\n<Subject 1> is the young male barber in <Picture 2>, preserving his short dark curly hair, neatly shaped beard, black barber shirt, focused expression, and calm professional manner. He works with gentle, sensitive, highly controlled movements and uses one small hand-held spray bottle, thin dainty precision scissors, a soft brush, and a small handheld mirror.\n\n<Subject 2> is the seagull in <Picture 1> and <Picture 2>, preserving its white-and-grey feathers, yellow beak, alert round eyes, and black barber cape. Its hair begins long, wild, tangled, and uneven like <Picture 1>, then gradually becomes the soft, neat, rounded, suave hairstyle shown in <Picture 2>. The seagull remains calm, comfortable, and completely unharmed. During the haircut it occasionally blinks and often keeps its eyes semi-closed as if enjoying the relaxing treatment.\n\n<Subject 3> is the modern dark barbershop from the reference images, featuring black walls, mirrors, professional lighting, barber tools, and a clean cinematic atmosphere.\n\nsummary:\n[reference generation] The target video is an exactly 18-second photorealistic comedic ASMR haircut sequence. <Subject 1> gently approaches <Subject 2>, lightly wets the messy hair using one spray bottle, and then slowly trims only the tips of the longer wet strands using thin dainty precision scissors. A tasteful time-lapse montage shows him moving around the chair and refining different sections little by little. He lightly re-wets the hair once more, uses only a very small subtle dusting of powder, then softly brushes the finished hairstyle until it looks smooth, bouncy, and suave, finally holding a mirror up for the seagull to admire the result.\n\nretention_analysis:\n<Subject 1> (appears throughout after entering): fully_preserved - retain the barber’s identity, hairstyle, beard, black clothing, delicate professional handling, and calm gentle behaviour.\n\n<Subject 2> (appears throughout): fully_preserved - retain the seagull’s identity, feathers, beak, eyes, cape, and natural anatomy while transforming only the hairstyle from wild and overgrown to neat, rounded, bouncy, and suave.\n\n<Subject 3> (appears throughout): fully_preserved - maintain the dark modern barbershop, mirrors, tools, lighting, and realistic depth.\n\ndetailed_description:\nThe target video uses a photorealistic cinematic ASMR style with shallow depth of field, crisp macro detail, natural professional barber movements, gentle comedy, and no dialogue or background music. Do not use electric clippers.\n\n[Shot 1] The video opens on <Subject 2> sitting alone beneath the black barber cape with extremely wild, long, uneven hair sticking out in every direction, matching <Picture 1>. The seagull blinks calmly. <Subject 1> then steps into frame behind the chair holding one small spray bottle and a pair of thin, delicate barber scissors.\n\n[Shot 2] At 00:02.000, cut to a tight close-up as <Subject 1> gently wets the hair using the single spray bottle. He gives two or three soft, separate squirts, each one light and controlled rather than harsh or excessive. Droplets settle into the hair, making it visibly damp and easier to shape. The seagull slowly lowers its eyelids, appearing relaxed.\n\n[Shot 3] At 00:04.000, cut to an ultra-close macro shot of the thin scissor tips. The barber trims only the very ends of the wet longer strands. The pace is slow, sensitive, and precise, like a real careful barber doing finishing work: snip... pause... snip... pause... snip. The blades are slim and dainty, never bulky. The camera stays tightly focused on the scissor tips, fingertips, wet strands, and tiny cut ends falling softly. The sound of the scissors cutting damp hair is crisp, intimate, and relaxing.\n\n[Shot 4] At 00:07.000, begin a smooth time-lapse montage. <Subject 1> moves naturally around both sides and behind the seagull, gently lifting sections, trimming tiny amounts from different areas, checking the shape, and refining the outline little by little. The seagull remains still with its eyes semi-closed, occasionally blinking as if enjoying the pampering. The messy hairstyle gradually becomes more balanced, rounded, and stylish.\n\n[Shot 5] At 00:11.000, during the time-lapse, the barber lightly re-wets the hair again with the same spray bottle using one or two soft squirts, then continues making tiny slow finishing snips. He never removes too much at once. The hairstyle becomes smooth, soft, and refined rather than short or overly cut.\n\n[Shot 6] At 00:13.500, return to normal speed in a close macro angle. <Subject 1> applies only a very small, subtle puff of barber powder, just enough to lightly dust the hair without creating a large cloud. He then gently brushes away any loose strands with a soft barber brush.\n\n[Shot 7] At 00:15.000, a close front angle shows the barber brushing and styling the hair with soft controlled strokes. The hair settles into a clean, rounded, suave shape with a soft bounce. The seagull’s eyes remain partially closed, calm and content.\n\n[Shot 8] At 00:17.000, <Subject 1> holds a small handheld mirror in front of <Subject 2>. The seagull looks at its reflection, opens its eyes, gives one slow satisfied blink, and subtly lifts its head with a proud expression. The barber smiles behind it as the finished hairstyle gives one final soft bouncy settle. End exactly at 00:18.000.\n\noverall_soundscape:\nDetailed close-range relaxing ASMR audio: soft trigger-sound spray bottle squirts from a single bottle, water settling into hair, extremely crisp slow scissor snips, the delicate sound of thin scissors cutting wet hair, faint hair strands falling, gentle barber hand movement, light brushing, a very soft minimal powder puff, feather rustling, cape fabric movement, quiet mirror handling, and subtle low barbershop room tone. Keep the audio intimate, soothing, and highly synchronized.\n\nnon_diegetic_music:\nN/A",
544
+ 1344,
545
+ 768,
546
+ 73
547
+ ]
548
+ },
549
+ {
550
+ "id": 146,
551
+ "type": "ImageScaleToTotalPixels",
552
+ "pos": [
553
+ -1942.2774418949214,
554
+ 4666.870619268523
555
+ ],
556
+ "size": [
557
+ 280,
558
+ 150
559
+ ],
560
+ "flags": {},
561
+ "order": 16,
562
+ "mode": 0,
563
+ "showAdvanced": true,
564
+ "inputs": [
565
+ {
566
+ "name": "image",
567
+ "type": "IMAGE",
568
+ "link": 268
569
+ }
570
+ ],
571
+ "outputs": [
572
+ {
573
+ "name": "IMAGE",
574
+ "type": "IMAGE",
575
+ "links": [
576
+ 269
577
+ ]
578
+ }
579
+ ],
580
+ "properties": {
581
+ "ue_properties": {
582
+ "widget_ue_connectable": {},
583
+ "version": "7.8",
584
+ "input_ue_unconnectable": {}
585
+ },
586
+ "cnr_id": "comfy-core",
587
+ "ver": "0.30.0",
588
+ "Node name for S&R": "ImageScaleToTotalPixels"
589
+ },
590
+ "widgets_values": [
591
+ "nearest-exact",
592
+ 1,
593
+ 32
594
+ ]
595
+ },
596
+ {
597
+ "id": 119,
598
+ "type": "ImageScaleToTotalPixels",
599
+ "pos": [
600
+ -2238.0886109920903,
601
+ 4670.573634714762
602
+ ],
603
+ "size": [
604
+ 280,
605
+ 150
606
+ ],
607
+ "flags": {},
608
+ "order": 15,
609
+ "mode": 0,
610
+ "showAdvanced": true,
611
+ "inputs": [
612
+ {
613
+ "name": "image",
614
+ "type": "IMAGE",
615
+ "link": 260
616
+ }
617
+ ],
618
+ "outputs": [
619
+ {
620
+ "name": "IMAGE",
621
+ "type": "IMAGE",
622
+ "links": [
623
+ 261
624
+ ]
625
+ }
626
+ ],
627
+ "properties": {
628
+ "cnr_id": "comfy-core",
629
+ "ver": "0.30.0",
630
+ "Node name for S&R": "ImageScaleToTotalPixels",
631
+ "ue_properties": {
632
+ "widget_ue_connectable": {},
633
+ "version": "7.8",
634
+ "input_ue_unconnectable": {}
635
+ }
636
+ },
637
+ "widgets_values": [
638
+ "nearest-exact",
639
+ 1,
640
+ 32
641
+ ]
642
+ },
643
+ {
644
+ "id": 117,
645
+ "type": "MarkdownNote",
646
+ "pos": [
647
+ -2766.004164312337,
648
+ 4292.809068762916
649
+ ],
650
+ "size": [
651
+ 440,
652
+ 680
653
+ ],
654
+ "flags": {},
655
+ "order": 3,
656
+ "mode": 0,
657
+ "inputs": [],
658
+ "outputs": [],
659
+ "title": "Note: Model Links",
660
+ "properties": {
661
+ "ue_properties": {
662
+ "widget_ue_connectable": {},
663
+ "version": "7.8",
664
+ "input_ue_unconnectable": {}
665
+ }
666
+ },
667
+ "widgets_values": [
668
+ "## Model Links\n\n**vae**\n\n- [minimax_h3_video_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors)\n- [minimax_h3_audio_vae_fp32.safetensors](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors)\n\n**diffusion_models**\n\n- [minimax_h3_fl2va_pruned_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_fl2va_pruned_int8_convrot.safetensors)\n\n**text_encoders**\n\n- [qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors)\n\n\n## Model Storage Location\n\n```\n📂 ComfyUI/\n├── 📂 models/\n│ ├── 📂 vae/\n│ │ ├── minimax_h3_video_vae_fp16.safetensors\n│ │ └── minimax_h3_audio_vae_fp32.safetensors\n│ ├── 📂 diffusion_models/\n│ │ └── minimax_h3_fl2va_pruned_int8_convrot.safetensors\n│ └── 📂 text_encoders/\n│ └── qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\n```\n\n## Report Issue\n\nNote: Please update ComfyUI first ([guide](https://docs.comfy.org/installation/update_comfyui)) and prepare required models. Desktop/Cloud updates follow stable releases, so some nightly-supported models may not be available yet.\n\n- Cannot run / runtime errors: [ComfyUI/issues](https://github.com/comfyanonymous/ComfyUI/issues)\n- UI / frontend issues: [ComfyUI_frontend/issues](https://github.com/Comfy-Org/ComfyUI_frontend/issues)\n- Workflow issues: [workflow_templates/issues](https://github.com/Comfy-Org/workflow_templates/issues)\n"
669
+ ],
670
+ "color": "#222",
671
+ "bgcolor": "#000"
672
+ },
673
+ {
674
+ "id": 116,
675
+ "type": "MarkdownNote",
676
+ "pos": [
677
+ -2759.5049670256058,
678
+ 5026.958844644791
679
+ ],
680
+ "size": [
681
+ 450,
682
+ 630
683
+ ],
684
+ "flags": {},
685
+ "order": 4,
686
+ "mode": 0,
687
+ "inputs": [],
688
+ "outputs": [],
689
+ "title": "Note: MiniMax H3",
690
+ "properties": {
691
+ "ue_properties": {
692
+ "widget_ue_connectable": {},
693
+ "version": "7.8",
694
+ "input_ue_unconnectable": {}
695
+ }
696
+ },
697
+ "widgets_values": [
698
+ "## MiniMax H3\n\n[MiniMax H3](https://www.minimax.io/blog/minimax-h3) is MiniMax's general-purpose, omni-modal generation model. It jointly understands text, image, video, and audio, and generates video with **native stereo audio**: voice, sound effects, and music are modeled jointly in a single forward pass, not layered on afterward. Output is up to 2K resolution, 24fps, and up to about 15 seconds.\n\n## About this workflow\n\nThis template runs the **Image to Video** task (`MiniMaxH3ImageToVideo` node), which covers both:\n\n- **t2va** (text-to-video), when no images are connected\n- **fl2va** (first/last-frame image-to-video), when `first_frame` and/or `last_frame` are connected\n\n**Key inputs**\n\n- **first_frame / last_frame**: optional keyframes; the model generates the motion between them\n- **prompt**: describe the shots, motion, and the accompanying audio (dialogue, SFX, music) in one block\n- **width / height**: set via Resolution Selector. H3's native canvas is a 768px short edge, capped at 768x1344 pixels, rounded to a multiple of 32\n- **duration (seconds)**: converted to a valid frame `length` by the Math Expression node, snapping up to the model's 17-frame-per-block (17k+5) grid at 24fps"
699
+ ],
700
+ "color": "#222",
701
+ "bgcolor": "#000"
702
+ },
703
+ {
704
+ "id": 137,
705
+ "type": "MiniMaxH3MemoryEfficientSageAttentionPatch",
706
+ "pos": [
707
+ -712.1886380294019,
708
+ 3692.3981147757795
709
+ ],
710
+ "size": [
711
+ 320.323828125,
712
+ 42.27849444257981
713
+ ],
714
+ "flags": {},
715
+ "order": 19,
716
+ "mode": 4,
717
+ "inputs": [
718
+ {
719
+ "name": "model",
720
+ "type": "MODEL",
721
+ "link": 251
722
+ }
723
+ ],
724
+ "outputs": [
725
+ {
726
+ "name": "model",
727
+ "type": "MODEL",
728
+ "links": [
729
+ 252
730
+ ]
731
+ }
732
+ ],
733
+ "properties": {
734
+ "cnr_id": "comfyui-kjnodes",
735
+ "ver": "1289b52fbb6d64a339a4047b9ea74cf7758ccf1e",
736
+ "Node name for S&R": "MiniMaxH3MemoryEfficientSageAttentionPatch",
737
+ "ue_properties": {
738
+ "widget_ue_connectable": {},
739
+ "input_ue_unconnectable": {},
740
+ "version": "7.8"
741
+ }
742
+ },
743
+ "widgets_values": []
744
+ },
745
+ {
746
+ "id": 140,
747
+ "type": "ModelPreviewOverrideKJ",
748
+ "pos": [
749
+ 55.0758538643357,
750
+ 4431.041713974029
751
+ ],
752
+ "size": [
753
+ 347.4377401781444,
754
+ 904.1019468088243
755
+ ],
756
+ "flags": {},
757
+ "order": 22,
758
+ "mode": 0,
759
+ "inputs": [
760
+ {
761
+ "name": "model",
762
+ "type": "MODEL",
763
+ "link": 254
764
+ },
765
+ {
766
+ "name": "vae",
767
+ "shape": 7,
768
+ "type": "VAE",
769
+ "link": null
770
+ }
771
+ ],
772
+ "outputs": [
773
+ {
774
+ "name": "MODEL",
775
+ "type": "MODEL",
776
+ "links": [
777
+ 263
778
+ ]
779
+ }
780
+ ],
781
+ "properties": {
782
+ "cnr_id": "comfyui-kjnodes",
783
+ "ver": "6edfa76599c13905968841b81309784a7dbb7803",
784
+ "Node name for S&R": "ModelPreviewOverrideKJ",
785
+ "ue_properties": {
786
+ "widget_ue_connectable": {},
787
+ "input_ue_unconnectable": {},
788
+ "version": "7.8"
789
+ }
790
+ },
791
+ "widgets_values": [
792
+ 1024,
793
+ 80,
794
+ true,
795
+ 360,
796
+ 8,
797
+ "taeh3.safetensors",
798
+ ""
799
+ ]
800
+ },
801
+ {
802
+ "id": 115,
803
+ "type": "ResolutionSelector",
804
+ "pos": [
805
+ -850.2751649697645,
806
+ 5397.131780627381
807
+ ],
808
+ "size": [
809
+ 270,
810
+ 170
811
+ ],
812
+ "flags": {},
813
+ "order": 5,
814
+ "mode": 0,
815
+ "showAdvanced": false,
816
+ "inputs": [],
817
+ "outputs": [
818
+ {
819
+ "name": "width",
820
+ "type": "INT",
821
+ "links": [
822
+ 246
823
+ ]
824
+ },
825
+ {
826
+ "name": "height",
827
+ "type": "INT",
828
+ "links": [
829
+ 247
830
+ ]
831
+ }
832
+ ],
833
+ "properties": {
834
+ "cnr_id": "comfy-core",
835
+ "ver": "0.30.0",
836
+ "Node name for S&R": "ResolutionSelector",
837
+ "ue_properties": {
838
+ "widget_ue_connectable": {},
839
+ "version": "7.8",
840
+ "input_ue_unconnectable": {}
841
+ }
842
+ },
843
+ "widgets_values": [
844
+ "9:16 (Portrait Widescreen)",
845
+ 1.2,
846
+ 32
847
+ ]
848
+ },
849
+ {
850
+ "id": 135,
851
+ "type": "PrimitiveFloat",
852
+ "pos": [
853
+ -1529.1560688912878,
854
+ 5277.633948952767
855
+ ],
856
+ "size": [
857
+ 270,
858
+ 70
859
+ ],
860
+ "flags": {},
861
+ "order": 6,
862
+ "mode": 0,
863
+ "inputs": [],
864
+ "outputs": [
865
+ {
866
+ "name": "FLOAT",
867
+ "type": "FLOAT",
868
+ "links": [
869
+ 248
870
+ ]
871
+ }
872
+ ],
873
+ "title": "Float (duration)",
874
+ "properties": {
875
+ "cnr_id": "comfy-core",
876
+ "ver": "0.30.0",
877
+ "Node name for S&R": "PrimitiveFloat",
878
+ "ue_properties": {
879
+ "widget_ue_connectable": {},
880
+ "version": "7.8",
881
+ "input_ue_unconnectable": {}
882
+ }
883
+ },
884
+ "widgets_values": [
885
+ 18
886
+ ]
887
+ },
888
+ {
889
+ "id": 122,
890
+ "type": "VAELoader",
891
+ "pos": [
892
+ -1569.0981259118,
893
+ 5140.776172362862
894
+ ],
895
+ "size": [
896
+ 650,
897
+ 70
898
+ ],
899
+ "flags": {},
900
+ "order": 7,
901
+ "mode": 0,
902
+ "inputs": [],
903
+ "outputs": [
904
+ {
905
+ "name": "VAE",
906
+ "type": "VAE",
907
+ "links": [
908
+ 229
909
+ ]
910
+ }
911
+ ],
912
+ "properties": {
913
+ "cnr_id": "comfy-core",
914
+ "ver": "0.30.0",
915
+ "Node name for S&R": "VAELoader",
916
+ "ue_properties": {
917
+ "widget_ue_connectable": {},
918
+ "version": "7.8",
919
+ "input_ue_unconnectable": {}
920
+ },
921
+ "models": [
922
+ {
923
+ "name": "minimax_h3_audio_vae_fp32.safetensors",
924
+ "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors",
925
+ "directory": "vae"
926
+ }
927
+ ]
928
+ },
929
+ "widgets_values": [
930
+ "minimax_h3_audio_vae_fp32.safetensors"
931
+ ]
932
+ },
933
+ {
934
+ "id": 121,
935
+ "type": "VAELoader",
936
+ "pos": [
937
+ -1569.0981259118,
938
+ 5003.81178772004
939
+ ],
940
+ "size": [
941
+ 640,
942
+ 70
943
+ ],
944
+ "flags": {},
945
+ "order": 8,
946
+ "mode": 0,
947
+ "inputs": [],
948
+ "outputs": [
949
+ {
950
+ "name": "VAE",
951
+ "type": "VAE",
952
+ "links": [
953
+ 230,
954
+ 242
955
+ ]
956
+ }
957
+ ],
958
+ "properties": {
959
+ "cnr_id": "comfy-core",
960
+ "ver": "0.30.0",
961
+ "Node name for S&R": "VAELoader",
962
+ "ue_properties": {
963
+ "widget_ue_connectable": {},
964
+ "version": "7.8",
965
+ "input_ue_unconnectable": {}
966
+ },
967
+ "models": [
968
+ {
969
+ "name": "minimax_h3_video_vae_fp16.safetensors",
970
+ "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors",
971
+ "directory": "vae"
972
+ }
973
+ ]
974
+ },
975
+ "widgets_values": [
976
+ "minimax_h3_video_vae_fp16.safetensors"
977
+ ]
978
+ },
979
+ {
980
+ "id": 142,
981
+ "type": "DiffusionModelLoaderKJ",
982
+ "pos": [
983
+ -1550.784989575604,
984
+ 4616.111271485676
985
+ ],
986
+ "size": [
987
+ 499.08960912716384,
988
+ 178
989
+ ],
990
+ "flags": {},
991
+ "order": 9,
992
+ "mode": 0,
993
+ "inputs": [
994
+ {
995
+ "name": "extra_state_dict",
996
+ "shape": 7,
997
+ "type": "STRING",
998
+ "link": null
999
+ }
1000
+ ],
1001
+ "outputs": [
1002
+ {
1003
+ "name": "MODEL",
1004
+ "type": "MODEL",
1005
+ "links": [
1006
+ 266
1007
+ ]
1008
+ }
1009
+ ],
1010
+ "properties": {
1011
+ "cnr_id": "comfyui-kjnodes",
1012
+ "ver": "dcfcb5dc9bad14912272cedd351cbb377bd27a22",
1013
+ "Node name for S&R": "DiffusionModelLoaderKJ",
1014
+ "ue_properties": {
1015
+ "widget_ue_connectable": {},
1016
+ "input_ue_unconnectable": {},
1017
+ "version": "7.8"
1018
+ }
1019
+ },
1020
+ "widgets_values": [
1021
+ "minimax_h3_fl2va_pruned_int8_convrot.safetensors",
1022
+ "default",
1023
+ "default",
1024
+ false,
1025
+ "disabled",
1026
+ false
1027
+ ]
1028
+ },
1029
+ {
1030
+ "id": 139,
1031
+ "type": "SpectrumApplyMiniMaxH3",
1032
+ "pos": [
1033
+ -79.05577752576175,
1034
+ 3649.89947617932
1035
+ ],
1036
+ "size": [
1037
+ 288.8521484375,
1038
+ 298
1039
+ ],
1040
+ "flags": {},
1041
+ "order": 21,
1042
+ "mode": 4,
1043
+ "inputs": [
1044
+ {
1045
+ "name": "model",
1046
+ "type": "MODEL",
1047
+ "link": 253
1048
+ }
1049
+ ],
1050
+ "outputs": [
1051
+ {
1052
+ "name": "model",
1053
+ "type": "MODEL",
1054
+ "links": [
1055
+ 254
1056
+ ]
1057
+ }
1058
+ ],
1059
+ "properties": {
1060
+ "cnr_id": "comfyui-spectrum-minimax-h3",
1061
+ "ver": "85ec1da66277e893079ecd46e32cc865c56cfe53",
1062
+ "Node name for S&R": "SpectrumApplyMiniMaxH3",
1063
+ "ue_properties": {
1064
+ "widget_ue_connectable": {},
1065
+ "input_ue_unconnectable": {},
1066
+ "version": "7.8"
1067
+ }
1068
+ },
1069
+ "widgets_values": [
1070
+ true,
1071
+ 0.5,
1072
+ 4,
1073
+ 0.1,
1074
+ 2,
1075
+ 0.75,
1076
+ 5,
1077
+ 1,
1078
+ 8,
1079
+ false,
1080
+ "system_ram"
1081
+ ]
1082
+ },
1083
+ {
1084
+ "id": 138,
1085
+ "type": "SolAttnPatch",
1086
+ "pos": [
1087
+ -368.81330214922974,
1088
+ 3663.2316283896935
1089
+ ],
1090
+ "size": [
1091
+ 270,
1092
+ 322
1093
+ ],
1094
+ "flags": {},
1095
+ "order": 20,
1096
+ "mode": 4,
1097
+ "inputs": [
1098
+ {
1099
+ "name": "model",
1100
+ "type": "MODEL",
1101
+ "link": 252
1102
+ }
1103
+ ],
1104
+ "outputs": [
1105
+ {
1106
+ "name": "MODEL",
1107
+ "type": "MODEL",
1108
+ "links": [
1109
+ 253
1110
+ ]
1111
+ }
1112
+ ],
1113
+ "properties": {
1114
+ "aux_id": "kijai/ComfyUI-SolAttn_triton",
1115
+ "ver": "1b8deced8297c51077d5625a5108e6f3f3d67093",
1116
+ "Node name for S&R": "SolAttnPatch",
1117
+ "ue_properties": {
1118
+ "widget_ue_connectable": {},
1119
+ "input_ue_unconnectable": {},
1120
+ "version": "7.8"
1121
+ }
1122
+ },
1123
+ "widgets_values": [
1124
+ 1.5,
1125
+ 0.2,
1126
+ 0.9,
1127
+ 4096,
1128
+ false,
1129
+ "exact_kv",
1130
+ false,
1131
+ "3d",
1132
+ false,
1133
+ false,
1134
+ false,
1135
+ ""
1136
+ ]
1137
+ },
1138
+ {
1139
+ "id": 136,
1140
+ "type": "PathchSageAttentionKJ",
1141
+ "pos": [
1142
+ -1002.770625682595,
1143
+ 3677.6514007243823
1144
+ ],
1145
+ "size": [
1146
+ 270,
1147
+ 82
1148
+ ],
1149
+ "flags": {},
1150
+ "order": 17,
1151
+ "mode": 4,
1152
+ "inputs": [
1153
+ {
1154
+ "name": "model",
1155
+ "type": "MODEL",
1156
+ "link": 267
1157
+ }
1158
+ ],
1159
+ "outputs": [
1160
+ {
1161
+ "name": "MODEL",
1162
+ "type": "MODEL",
1163
+ "links": [
1164
+ 251
1165
+ ]
1166
+ }
1167
+ ],
1168
+ "properties": {
1169
+ "cnr_id": "comfyui-kjnodes",
1170
+ "ver": "1289b52fbb6d64a339a4047b9ea74cf7758ccf1e",
1171
+ "Node name for S&R": "PathchSageAttentionKJ",
1172
+ "ue_properties": {
1173
+ "widget_ue_connectable": {},
1174
+ "input_ue_unconnectable": {},
1175
+ "version": "7.8"
1176
+ }
1177
+ },
1178
+ "widgets_values": [
1179
+ "auto",
1180
+ false
1181
+ ]
1182
+ },
1183
+ {
1184
+ "id": 118,
1185
+ "type": "MarkdownNote",
1186
+ "pos": [
1187
+ -1618.7340725056556,
1188
+ 5415.983577320917
1189
+ ],
1190
+ "size": [
1191
+ 300,
1192
+ 520
1193
+ ],
1194
+ "flags": {},
1195
+ "order": 10,
1196
+ "mode": 0,
1197
+ "inputs": [],
1198
+ "outputs": [],
1199
+ "title": "Note: Size Settings Reference",
1200
+ "properties": {
1201
+ "ue_properties": {
1202
+ "widget_ue_connectable": {},
1203
+ "version": "7.8",
1204
+ "input_ue_unconnectable": {}
1205
+ }
1206
+ },
1207
+ "widgets_values": [
1208
+ "| megapixels | Aspect | Output (multiple=32) |\n|---|---|---|\n| 0.2 | 16:9 | 608 x 352 |\n| 0.3 | 16:9 | 736 x 416 |\n| 0.4 | 16:9 | 864 x 480 |\n| 0.5 | 16:9 | 960 x 544 |\n| 0.6 | 16:9 | 1056 x 608 |\n| 0.7 | 16:9 | 1152 x 640 |\n| 0.8 | 16:9 | 1216 x 672 |\n| 0.9 | 16:9 | 1280 x 736 |\n| 0.98 | 16:9 | 1344 x 768 |\n| 1.0 | 16:9 | 1376 x 768 |\n| 1.2 | 16:9 | 1504 x 832 |\n| 1.5 | 16:9 | 1664 x 928 |\n| 1.8 | 16:9 | 1824 x 1024 |\n| 2.0 | 16:9 | 1920 x 1088 |\n"
1209
+ ],
1210
+ "color": "#222",
1211
+ "bgcolor": "#000"
1212
+ },
1213
+ {
1214
+ "id": 143,
1215
+ "type": "MiniMaxH3SigmaShift",
1216
+ "pos": [
1217
+ -350.83391503803983,
1218
+ 4400.461389732062
1219
+ ],
1220
+ "size": [
1221
+ 270,
1222
+ 82
1223
+ ],
1224
+ "flags": {},
1225
+ "order": 23,
1226
+ "mode": 0,
1227
+ "inputs": [
1228
+ {
1229
+ "name": "model",
1230
+ "type": "MODEL",
1231
+ "link": 263
1232
+ }
1233
+ ],
1234
+ "outputs": [
1235
+ {
1236
+ "name": "MODEL",
1237
+ "type": "MODEL",
1238
+ "links": [
1239
+ 264,
1240
+ 265
1241
+ ]
1242
+ }
1243
+ ],
1244
+ "properties": {
1245
+ "ue_properties": {
1246
+ "widget_ue_connectable": {},
1247
+ "input_ue_unconnectable": {}
1248
+ },
1249
+ "cnr_id": "comfy-core",
1250
+ "ver": "0.30.0",
1251
+ "Node name for S&R": "MiniMaxH3SigmaShift"
1252
+ },
1253
+ "widgets_values": [
1254
+ 12,
1255
+ 6
1256
+ ]
1257
+ },
1258
+ {
1259
+ "id": 141,
1260
+ "type": "VHS_VideoCombine",
1261
+ "pos": [
1262
+ 714.1266838626509,
1263
+ 4518.4007305176365
1264
+ ],
1265
+ "size": [
1266
+ 711.6150662080554,
1267
+ 1598.2272350684077
1268
+ ],
1269
+ "flags": {},
1270
+ "order": 29,
1271
+ "mode": 0,
1272
+ "inputs": [
1273
+ {
1274
+ "name": "images",
1275
+ "type": "IMAGE",
1276
+ "link": 255
1277
+ },
1278
+ {
1279
+ "name": "audio",
1280
+ "shape": 7,
1281
+ "type": "AUDIO",
1282
+ "link": 256
1283
+ },
1284
+ {
1285
+ "name": "meta_batch",
1286
+ "shape": 7,
1287
+ "type": "VHS_BatchManager",
1288
+ "link": null
1289
+ },
1290
+ {
1291
+ "name": "vae",
1292
+ "shape": 7,
1293
+ "type": "VAE",
1294
+ "link": null
1295
+ }
1296
+ ],
1297
+ "outputs": [
1298
+ {
1299
+ "name": "Filenames",
1300
+ "type": "VHS_FILENAMES",
1301
+ "links": null
1302
+ }
1303
+ ],
1304
+ "properties": {
1305
+ "cnr_id": "comfyui-videohelpersuite",
1306
+ "ver": "4ee72c065db22c9d96c2427954dc69e7b908444b",
1307
+ "Node name for S&R": "VHS_VideoCombine",
1308
+ "ue_properties": {
1309
+ "widget_ue_connectable": {},
1310
+ "input_ue_unconnectable": {},
1311
+ "version": "7.8"
1312
+ }
1313
+ },
1314
+ "widgets_values": {
1315
+ "frame_rate": 24,
1316
+ "loop_count": 0,
1317
+ "filename_prefix": "video/MiniMax_H3",
1318
+ "format": "video/h264-mp4",
1319
+ "pix_fmt": "yuv420p",
1320
+ "crf": 20,
1321
+ "save_metadata": true,
1322
+ "trim_to_audio": false,
1323
+ "pingpong": false,
1324
+ "save_output": true,
1325
+ "videopreview": {
1326
+ "hidden": false,
1327
+ "paused": false,
1328
+ "params": {
1329
+ "filename": "MiniMax_H3_00060-audio.mp4",
1330
+ "subfolder": "video",
1331
+ "type": "output",
1332
+ "format": "video/h264-mp4",
1333
+ "frame_rate": 24,
1334
+ "workflow": "MiniMax_H3_00060.png",
1335
+ "fullpath": "C:\\Users\\drbaph\\Documents\\ComfyUI\\output\\video\\MiniMax_H3_00060-audio.mp4"
1336
+ }
1337
+ }
1338
+ }
1339
+ },
1340
+ {
1341
+ "id": 144,
1342
+ "type": "LoraLoaderModelOnly",
1343
+ "pos": [
1344
+ -1170.925848542515,
1345
+ 4456.049086376686
1346
+ ],
1347
+ "size": [
1348
+ 270,
1349
+ 82
1350
+ ],
1351
+ "flags": {},
1352
+ "order": 14,
1353
+ "mode": 0,
1354
+ "inputs": [
1355
+ {
1356
+ "name": "model",
1357
+ "type": "MODEL",
1358
+ "link": 266
1359
+ }
1360
+ ],
1361
+ "outputs": [
1362
+ {
1363
+ "name": "MODEL",
1364
+ "type": "MODEL",
1365
+ "links": [
1366
+ 267
1367
+ ]
1368
+ }
1369
+ ],
1370
+ "properties": {
1371
+ "ue_properties": {
1372
+ "widget_ue_connectable": {},
1373
+ "input_ue_unconnectable": {}
1374
+ },
1375
+ "cnr_id": "comfy-core",
1376
+ "ver": "0.30.0",
1377
+ "Node name for S&R": "LoraLoaderModelOnly"
1378
+ },
1379
+ "widgets_values": [
1380
+ "h3\\minimax_h3_turbo_4step_comfyui.safetensors",
1381
+ 1
1382
+ ]
1383
+ },
1384
+ {
1385
+ "id": 114,
1386
+ "type": "LoadImage",
1387
+ "pos": [
1388
+ -2170.723500126082,
1389
+ 3951.2530874294207
1390
+ ],
1391
+ "size": [
1392
+ 450,
1393
+ 630
1394
+ ],
1395
+ "flags": {},
1396
+ "order": 11,
1397
+ "mode": 0,
1398
+ "inputs": [],
1399
+ "outputs": [
1400
+ {
1401
+ "name": "IMAGE",
1402
+ "type": "IMAGE",
1403
+ "links": [
1404
+ 260
1405
+ ]
1406
+ },
1407
+ {
1408
+ "name": "MASK",
1409
+ "type": "MASK",
1410
+ "links": null
1411
+ }
1412
+ ],
1413
+ "properties": {
1414
+ "cnr_id": "comfy-core",
1415
+ "ver": "0.30.0",
1416
+ "Node name for S&R": "LoadImage",
1417
+ "ue_properties": {
1418
+ "widget_ue_connectable": {},
1419
+ "version": "7.8",
1420
+ "input_ue_unconnectable": {}
1421
+ }
1422
+ },
1423
+ "widgets_values": [
1424
+ "messy_hair_seagull.webp",
1425
+ "image"
1426
+ ]
1427
+ },
1428
+ {
1429
+ "id": 145,
1430
+ "type": "LoadImage",
1431
+ "pos": [
1432
+ -2172.76204891036,
1433
+ 4886.137863116034
1434
+ ],
1435
+ "size": [
1436
+ 450,
1437
+ 630
1438
+ ],
1439
+ "flags": {},
1440
+ "order": 12,
1441
+ "mode": 0,
1442
+ "inputs": [],
1443
+ "outputs": [
1444
+ {
1445
+ "name": "IMAGE",
1446
+ "type": "IMAGE",
1447
+ "links": [
1448
+ 268
1449
+ ]
1450
+ },
1451
+ {
1452
+ "name": "MASK",
1453
+ "type": "MASK",
1454
+ "links": null
1455
+ }
1456
+ ],
1457
+ "properties": {
1458
+ "ue_properties": {
1459
+ "widget_ue_connectable": {},
1460
+ "version": "7.8",
1461
+ "input_ue_unconnectable": {}
1462
+ },
1463
+ "cnr_id": "comfy-core",
1464
+ "ver": "0.30.0",
1465
+ "Node name for S&R": "LoadImage"
1466
+ },
1467
+ "widgets_values": [
1468
+ "barber_seagull_haircut.webp",
1469
+ "image"
1470
+ ]
1471
+ }
1472
+ ],
1473
+ "links": [
1474
+ [
1475
+ 229,
1476
+ 122,
1477
+ 0,
1478
+ 123,
1479
+ 1,
1480
+ "VAE"
1481
+ ],
1482
+ [
1483
+ 230,
1484
+ 121,
1485
+ 0,
1486
+ 124,
1487
+ 1,
1488
+ "VAE"
1489
+ ],
1490
+ [
1491
+ 232,
1492
+ 131,
1493
+ 0,
1494
+ 127,
1495
+ 0,
1496
+ "NOISE"
1497
+ ],
1498
+ [
1499
+ 233,
1500
+ 128,
1501
+ 0,
1502
+ 127,
1503
+ 1,
1504
+ "GUIDER"
1505
+ ],
1506
+ [
1507
+ 234,
1508
+ 125,
1509
+ 0,
1510
+ 127,
1511
+ 2,
1512
+ "SAMPLER"
1513
+ ],
1514
+ [
1515
+ 235,
1516
+ 126,
1517
+ 0,
1518
+ 127,
1519
+ 3,
1520
+ "SIGMAS"
1521
+ ],
1522
+ [
1523
+ 236,
1524
+ 133,
1525
+ 1,
1526
+ 127,
1527
+ 4,
1528
+ "LATENT"
1529
+ ],
1530
+ [
1531
+ 238,
1532
+ 133,
1533
+ 0,
1534
+ 128,
1535
+ 1,
1536
+ "CONDITIONING"
1537
+ ],
1538
+ [
1539
+ 241,
1540
+ 130,
1541
+ 0,
1542
+ 133,
1543
+ 0,
1544
+ "CLIP"
1545
+ ],
1546
+ [
1547
+ 242,
1548
+ 121,
1549
+ 0,
1550
+ 133,
1551
+ 1,
1552
+ "VAE"
1553
+ ],
1554
+ [
1555
+ 245,
1556
+ 134,
1557
+ 1,
1558
+ 133,
1559
+ 6,
1560
+ "INT"
1561
+ ],
1562
+ [
1563
+ 246,
1564
+ 115,
1565
+ 0,
1566
+ 133,
1567
+ 4,
1568
+ "INT"
1569
+ ],
1570
+ [
1571
+ 247,
1572
+ 115,
1573
+ 1,
1574
+ 133,
1575
+ 5,
1576
+ "INT"
1577
+ ],
1578
+ [
1579
+ 248,
1580
+ 135,
1581
+ 0,
1582
+ 134,
1583
+ 0,
1584
+ "FLOAT"
1585
+ ],
1586
+ [
1587
+ 249,
1588
+ 127,
1589
+ 0,
1590
+ 124,
1591
+ 0,
1592
+ "LATENT"
1593
+ ],
1594
+ [
1595
+ 250,
1596
+ 127,
1597
+ 0,
1598
+ 123,
1599
+ 0,
1600
+ "LATENT"
1601
+ ],
1602
+ [
1603
+ 251,
1604
+ 136,
1605
+ 0,
1606
+ 137,
1607
+ 0,
1608
+ "MODEL"
1609
+ ],
1610
+ [
1611
+ 252,
1612
+ 137,
1613
+ 0,
1614
+ 138,
1615
+ 0,
1616
+ "MODEL"
1617
+ ],
1618
+ [
1619
+ 253,
1620
+ 138,
1621
+ 0,
1622
+ 139,
1623
+ 0,
1624
+ "MODEL"
1625
+ ],
1626
+ [
1627
+ 254,
1628
+ 139,
1629
+ 0,
1630
+ 140,
1631
+ 0,
1632
+ "MODEL"
1633
+ ],
1634
+ [
1635
+ 255,
1636
+ 124,
1637
+ 0,
1638
+ 141,
1639
+ 0,
1640
+ "IMAGE"
1641
+ ],
1642
+ [
1643
+ 256,
1644
+ 123,
1645
+ 0,
1646
+ 141,
1647
+ 1,
1648
+ "AUDIO"
1649
+ ],
1650
+ [
1651
+ 260,
1652
+ 114,
1653
+ 0,
1654
+ 119,
1655
+ 0,
1656
+ "IMAGE"
1657
+ ],
1658
+ [
1659
+ 261,
1660
+ 119,
1661
+ 0,
1662
+ 133,
1663
+ 2,
1664
+ "IMAGE"
1665
+ ],
1666
+ [
1667
+ 263,
1668
+ 140,
1669
+ 0,
1670
+ 143,
1671
+ 0,
1672
+ "MODEL"
1673
+ ],
1674
+ [
1675
+ 264,
1676
+ 143,
1677
+ 0,
1678
+ 128,
1679
+ 0,
1680
+ "MODEL"
1681
+ ],
1682
+ [
1683
+ 265,
1684
+ 143,
1685
+ 0,
1686
+ 126,
1687
+ 0,
1688
+ "MODEL"
1689
+ ],
1690
+ [
1691
+ 266,
1692
+ 142,
1693
+ 0,
1694
+ 144,
1695
+ 0,
1696
+ "MODEL"
1697
+ ],
1698
+ [
1699
+ 267,
1700
+ 144,
1701
+ 0,
1702
+ 136,
1703
+ 0,
1704
+ "MODEL"
1705
+ ],
1706
+ [
1707
+ 268,
1708
+ 145,
1709
+ 0,
1710
+ 146,
1711
+ 0,
1712
+ "IMAGE"
1713
+ ],
1714
+ [
1715
+ 269,
1716
+ 146,
1717
+ 0,
1718
+ 133,
1719
+ 3,
1720
+ "IMAGE"
1721
+ ]
1722
+ ],
1723
+ "groups": [
1724
+ {
1725
+ "id": 5,
1726
+ "title": "Use Image Size",
1727
+ "bounding": [
1728
+ -859.1290060636652,
1729
+ 5307.775425175882,
1730
+ 296.484375,
1731
+ 354
1732
+ ],
1733
+ "color": "#3f789e",
1734
+ "flags": {}
1735
+ },
1736
+ {
1737
+ "id": 1,
1738
+ "title": "Models",
1739
+ "bounding": [
1740
+ -1600.615933590389,
1741
+ 4546.578902898876,
1742
+ 700,
1743
+ 800
1744
+ ],
1745
+ "color": "#3f789e",
1746
+ "flags": {}
1747
+ },
1748
+ {
1749
+ "id": 2,
1750
+ "title": "Sampling",
1751
+ "bounding": [
1752
+ -365.1773390857757,
1753
+ 4487.280631458884,
1754
+ 700,
1755
+ 800
1756
+ ],
1757
+ "color": "#3f789e",
1758
+ "flags": {}
1759
+ },
1760
+ {
1761
+ "id": 3,
1762
+ "title": "Conditioning",
1763
+ "bounding": [
1764
+ -875.1773390857762,
1765
+ 4487.280631458884,
1766
+ 480,
1767
+ 800
1768
+ ],
1769
+ "color": "#3f789e",
1770
+ "flags": {}
1771
+ },
1772
+ {
1773
+ "id": 4,
1774
+ "title": "Decoding and create video",
1775
+ "bounding": [
1776
+ 354.8226609142243,
1777
+ 4487.280631458884,
1778
+ 700,
1779
+ 800
1780
+ ],
1781
+ "color": "#3f789e",
1782
+ "flags": {}
1783
+ },
1784
+ {
1785
+ "id": 6,
1786
+ "title": "Optimisations (optional)",
1787
+ "bounding": [
1788
+ -1018.8418563969516,
1789
+ 3579.89947617932,
1790
+ 1238.638227308687,
1791
+ 439.6170750677993
1792
+ ],
1793
+ "color": "#3f789e",
1794
+ "flags": {}
1795
+ }
1796
+ ],
1797
+ "config": {},
1798
+ "extra": {
1799
+ "frontendVersion": "1.48.6",
1800
+ "VHS_latentpreview": false,
1801
+ "VHS_latentpreviewrate": 0,
1802
+ "VHS_MetadataImage": true,
1803
+ "VHS_KeepIntermediate": true,
1804
+ "ds": {
1805
+ "scale": 0.3380916604057227,
1806
+ "offset": [
1807
+ 3594.664231121092,
1808
+ -2846.4337842300465
1809
+ ]
1810
+ },
1811
+ "ue_links": [],
1812
+ "links_added_by_ue": []
1813
+ },
1814
+ "version": 0.4
1815
+ }
minimax_h3_turbo_4step_ckpt500_pruned_comfyui.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f92761a2e3c8ce9c61177eb3e8836f2bbcb937a6c1615389355cf6dd5dc960d
3
+ size 620285648
minimax_h3_turbo_4step_ema_ckpt500_pruned_comfyui.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ad5c813e23c8575bbfa16e7404c0c7f5ef274fb68665f61c860366b3ddd7211
3
+ size 620285648
minimax_h3_turbo_4step_ema_pruned_comfyui.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ad7ac0a3dc5b464f21a1cf1996650e5887be3405a178851efc5e103e337fe38
3
+ size 620285648
minimax_h3_turbo_4step_pruned_comfyui.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac0d2e6aca12cf3b6e12c5ca75d97aaafd02efd0943288e5285d74fd8f5f1277
3
+ size 620285648