ComfyPilotComfyPilot
工作流 / 视频 / FastVideo FastH3:图生视频
FastVideo FastH3:图生视频

FastVideo FastH3:图生视频

FastVideo FastH3: Image to Video

使用 MiniMax H3 和 FastH3 8 步检查点,从单张静态图像生成带有同步原生音频的视频。一张图像即可一次性输出一个视频,并配有匹配的语音、音效和音乐。非常适合让肖像动起来、让产品图栩栩如生,以及快速制作社交媒体短片。

视频 7 个节点 1 个输入 2,067 次使用 ComfyUI ≥ 0.36.0 ComfyUI 67 KB

下载工作流 JSON 查看官方来源

下载后把文件拖进 ComfyUI 窗口即可打开。

需要的权重

这个工作流引用了 4 个权重文件,其中 3 个能在本站模型库里找到。

fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors 模型库暂未收录
minimax_h3_audio_vae_fp32.safetensors 在模型库查看
minimax_h3_video_vae_int8_convrot.safetensors 在模型库查看
qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors 在模型库查看

节点构成

MarkdownNote × 2
SaveVideo × 1
4c314f31-ecda-4b08-ae98-faaba1bf613f × 1
LoadImage × 1
GetImageSize × 1
ImageScaleToTotalPixels × 1

官方标注的模型

MiniMax H3

工作流 JSON

和上面「下载工作流 JSON」是同一份文件。

1{
2 "id": "e3f2b845-8f2c-4b5a-9caf-eac1029d3e7e",
3 "revision": 0,
4 "last_node_id": 143,
5 "last_link_id": 274,
6 "nodes": [
7 {
8 "id": 92,
9 "type": "SaveVideo",
10 "pos": [
11 -370,
12 4880
13 ],
14 "size": [
15 940,
16 126
17 ],
18 "flags": [],
19 "order": 6,
20 "mode": 0,
21 "inputs": [
22 {
23 "name": "video",
24 "type": "VIDEO",
25 "link": 194
26 }
27 ],
28 "outputs": [
29 {
30 "name": "video",
31 "type": "VIDEO",
32 "links": null
33 }
34 ],
35 "properties": {
36 "cnr_id": "comfy-core",
37 "ver": "0.33.0"
38 },
39 "widgets_values": [
40 "video/MiniMax_H3",
41 "auto",
42 "auto",
43 "auto"
44 ],
45 "widgets_values_named": {
46 "filename_prefix": "video/MiniMax_H3",
47 "format": "auto",
48 "format.codec": "auto",
49 "codec": "auto"
50 }
51 },
52 {
53 "id": 105,
54 "type": "4c314f31-ecda-4b08-ae98-faaba1bf613f",
55 "pos": [
56 -900,
57 4880
58 ],
59 "size": [
60 480,
61 630
62 ],
63 "flags": [],
64 "order": 5,
65 "mode": 0,
66 "inputs": [
67 {
68 "name": "first_frame",
69 "shape": 7,
70 "type": "IMAGE",
71 "link": 262
72 },
73 {
74 "name": "last_frame",
75 "shape": 7,
76 "type": "IMAGE",
77 "link": null
78 },
79 {
80 "name": "width",
81 "type": "INT",
82 "widget": {
83 "name": "width"
84 },
85 "link": 271
86 },
87 {
88 "name": "height",
89 "type": "INT",
90 "widget": {
91 "name": "height"
92 },
93 "link": 272
94 },
95 {
96 "label": "duration",
97 "name": "value_1",
98 "type": "FLOAT",
99 "widget": {
100 "name": "value_1"
101 },
102 "link": null
103 },
104 {
105 "label": "audio_vae",
106 "name": "vae_name_1",
107 "type": "COMBO",
108 "widget": {
109 "name": "vae_name_1"
110 },
111 "link": null
112 }
113 ],
114 "outputs": [
115 {
116 "name": "VIDEO",
117 "type": "VIDEO",
118 "links": [
119 194
120 ]
121 }
122 ],
123 "properties": {
124 "cnr_id": "comfy-core",
125 "ver": "0.33.0",
126 "previewExposures": []
127 },
128 "widgets_values": [
129 "For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.\nintegrated_multimodal_description: [Shot 1] This is a live-action, highly stylized fashion video with a stark, high-contrast aesthetic and digital glitch overlays. The figure stands dead center against a plain off-white background that has a subtle, paper-like texture. The outfit is a layered, monochrome streetwear ensemble. The upper body is enveloped in a bulky, matte black puffer jacket with thick horizontal quilting and a central zipper, worn over a dark inner layer. The hands are tucked casually into the lower side pockets of the jacket. The lower body features loose-fitting, baggy black cargo pants with large side pockets, prominent seams, and a long, thin black fabric strap dangling freely from the left thigh area. The feet are clad in chunky, textured grey sneakers with intricate mesh detailing, thick soles, and dark grey laces tied neatly. The figure's head is completely concealed, creating an anonymous, edgy vibe. A black, textured hood is pulled up tightly, framing a black fabric face covering that reaches up to the bridge of the nose. Over the eyes, a pair of large, prominent round goggles with thick black frames and dark, opaque lenses completes the look. Cutting horizontally across the figure's midsection is a striking digital glitch effect. It consists of a thick, vibrant, solid red horizontal band, flanked by jagged, horizontal black streaks that blur and distort the fabric of the jacket behind it. The camera slowly pushes in on the static figure. As the camera moves, a low, synthesized bass drone is heard, and the thick red glitch band flickers violently, its edges stuttering horizontally, while the figure maintains a rigid, confident, and imposing stance, entirely motionless except for the digital distortion. [Shot 2] At 01.200, the camera cuts to a tight, dynamic close-up of the figure's head and upper shoulders, framed from a slightly low angle to emphasize dominance. The intricate textures of the headgear are now highly visible: the coarse weave of the black hood, the fine ribbing of the black face covering, and the smooth, matte finish of the thick black goggle frames. The dark, opaque lenses of the round goggles reflect a faint, unseen rectangular light source, adding depth. The camera executes a fast tilt-up motion. During this rapid upward movement, the figure slowly tilts their head downward, creating a contrasting motion. Simultaneously, a sharp, high-frequency digital stutter is heard as a secondary, localized digital glitch—a rapid sequence of thin, horizontal red and black scanlines—flashes across the reflective surface of the goggles for a split second. The bulky collar of the black puffer jacket frames the lower edge of the shot, remaining completely still. [Shot 3] At 02.500, the camera cuts to a medium close-up focusing specifically on the figure's lower torso, the baggy black cargo pants, and the hanging strap. The camera trucks right at a fast speed, creating a sense of lateral momentum. In response to the camera's movement, the figure shifts their weight from the left leg to the right leg. A heavy fabric rustle is heard as this subtle action causes the stiff fabric of the cargo pants to buckle and crease, creating deep shadows in the folds, while the long black fabric strap on the left thigh swings like a pendulum. As the strap swings, a new, jagged digital glitch effect abruptly tears across the left side of the frame. This glitch consists of sharp, alternating black and bright white horizontal lines that momentarily obscure the swinging strap and the side pocket of the pants, mimicking a corrupted video feed. The plain off-white background slides smoothly past in the background due to the trucking motion. [Shot 4] At 03.800, the camera cuts to a medium-wide shot, returning to a frontal perspective similar to the initial framing, but from a slightly closer vantage point. The camera zooms in quickly with a sudden, aggressive burst of speed. Reacting to the escalating energy, the figure abruptly pulls both hands out of the puffy black jacket pockets in a sharp, synchronized motion, and crosses their arms tightly over their chest. The fabric of the puffer jacket rustles and compresses. A loud, aggressive digital distortion crash is heard the moment the arms cross, and the primary thick red glitch band from Shot 1 reappears across the midsection. This time, the red band pulsates intensely, expanding vertically and horizontally, its edges fraying into chaotic, pixelated red and black blocks that briefly consume the figure's crossed arms and chest. The stark off-white background remains perfectly undisturbed, providing a clean canvas that highlights the aggressive digital distortion and the figure's final, striking pose as the clip concludes.\n\noverall_soundscape: The soundscape begins with a continuous, low-frequency synthesized bass drone that establishes a tense, modern atmosphere. As the first digital glitch flickers, a series of rapid, electrical crackling sounds and low-pitched digital stutters become clearly heard in the foreground. The moment the camera cuts to the close-up of the head, a sharp, high-frequency digital stutter cuts through the mix, synchronized with the visual flash on the goggles. Followed by this, a distinct, heavy fabric rustle is distinctly audible as the figure shifts their weight, the stiff material of the cargo pants rubbing together. Toward the end of the clip, as the figure sharply crosses their arms, a loud, aggressive digital distortion crash dominates the foreground, resembling a corrupted audio file tearing apart. This crash transitions into a final, heavy synthesized sub-bass impact that lingers as the video ends. A faint, rhythmic electronic clicking, like a fast metronome, plays consistently in the background throughout the entire sequence, tying the dynamic cuts together.\n\nnon_diegetic_music: A fast-paced, aggressive cyberpunk-style electronic track drives the sequence, featuring a heavy, distorted 808 bassline and rapid, syncopated hi-hat trills. The music maintains a high-energy, rhythmic pulse that perfectly matches the dynamic, multi-shot fashion edit and the digital glitch aesthetic.",
130 1344,
131 768,
132 5,
133 793369680225544,
134 "fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors",
135 "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
136 "minimax_h3_video_vae_int8_convrot.safetensors",
137 "minimax_h3_audio_vae_fp32.safetensors",
138 "comfy kitchen attention"
139 ],
140 "widgets_values_named": {
141 "prompt": "For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.\nintegrated_multimodal_description: [Shot 1] This is a live-action, highly stylized fashion video with a stark, high-contrast aesthetic and digital glitch overlays. The figure stands dead center against a plain off-white background that has a subtle, paper-like texture. The outfit is a layered, monochrome streetwear ensemble. The upper body is enveloped in a bulky, matte black puffer jacket with thick horizontal quilting and a central zipper, worn over a dark inner layer. The hands are tucked casually into the lower side pockets of the jacket. The lower body features loose-fitting, baggy black cargo pants with large side pockets, prominent seams, and a long, thin black fabric strap dangling freely from the left thigh area. The feet are clad in chunky, textured grey sneakers with intricate mesh detailing, thick soles, and dark grey laces tied neatly. The figure's head is completely concealed, creating an anonymous, edgy vibe. A black, textured hood is pulled up tightly, framing a black fabric face covering that reaches up to the bridge of the nose. Over the eyes, a pair of large, prominent round goggles with thick black frames and dark, opaque lenses completes the look. Cutting horizontally across the figure's midsection is a striking digital glitch effect. It consists of a thick, vibrant, solid red horizontal band, flanked by jagged, horizontal black streaks that blur and distort the fabric of the jacket behind it. The camera slowly pushes in on the static figure. As the camera moves, a low, synthesized bass drone is heard, and the thick red glitch band flickers violently, its edges stuttering horizontally, while the figure maintains a rigid, confident, and imposing stance, entirely motionless except for the digital distortion. [Shot 2] At 01.200, the camera cuts to a tight, dynamic close-up of the figure's head and upper shoulders, framed from a slightly low angle to emphasize dominance. The intricate textures of the headgear are now highly visible: the coarse weave of the black hood, the fine ribbing of the black face covering, and the smooth, matte finish of the thick black goggle frames. The dark, opaque lenses of the round goggles reflect a faint, unseen rectangular light source, adding depth. The camera executes a fast tilt-up motion. During this rapid upward movement, the figure slowly tilts their head downward, creating a contrasting motion. Simultaneously, a sharp, high-frequency digital stutter is heard as a secondary, localized digital glitch—a rapid sequence of thin, horizontal red and black scanlines—flashes across the reflective surface of the goggles for a split second. The bulky collar of the black puffer jacket frames the lower edge of the shot, remaining completely still. [Shot 3] At 02.500, the camera cuts to a medium close-up focusing specifically on the figure's lower torso, the baggy black cargo pants, and the hanging strap. The camera trucks right at a fast speed, creating a sense of lateral momentum. In response to the camera's movement, the figure shifts their weight from the left leg to the right leg. A heavy fabric rustle is heard as this subtle action causes the stiff fabric of the cargo pants to buckle and crease, creating deep shadows in the folds, while the long black fabric strap on the left thigh swings like a pendulum. As the strap swings, a new, jagged digital glitch effect abruptly tears across the left side of the frame. This glitch consists of sharp, alternating black and bright white horizontal lines that momentarily obscure the swinging strap and the side pocket of the pants, mimicking a corrupted video feed. The plain off-white background slides smoothly past in the background due to the trucking motion. [Shot 4] At 03.800, the camera cuts to a medium-wide shot, returning to a frontal perspective similar to the initial framing, but from a slightly closer vantage point. The camera zooms in quickly with a sudden, aggressive burst of speed. Reacting to the escalating energy, the figure abruptly pulls both hands out of the puffy black jacket pockets in a sharp, synchronized motion, and crosses their arms tightly over their chest. The fabric of the puffer jacket rustles and compresses. A loud, aggressive digital distortion crash is heard the moment the arms cross, and the primary thick red glitch band from Shot 1 reappears across the midsection. This time, the red band pulsates intensely, expanding vertically and horizontally, its edges fraying into chaotic, pixelated red and black blocks that briefly consume the figure's crossed arms and chest. The stark off-white background remains perfectly undisturbed, providing a clean canvas that highlights the aggressive digital distortion and the figure's final, striking pose as the clip concludes.\n\noverall_soundscape: The soundscape begins with a continuous, low-frequency synthesized bass drone that establishes a tense, modern atmosphere. As the first digital glitch flickers, a series of rapid, electrical crackling sounds and low-pitched digital stutters become clearly heard in the foreground. The moment the camera cuts to the close-up of the head, a sharp, high-frequency digital stutter cuts through the mix, synchronized with the visual flash on the goggles. Followed by this, a distinct, heavy fabric rustle is distinctly audible as the figure shifts their weight, the stiff material of the cargo pants rubbing together. Toward the end of the clip, as the figure sharply crosses their arms, a loud, aggressive digital distortion crash dominates the foreground, resembling a corrupted audio file tearing apart. This crash transitions into a final, heavy synthesized sub-bass impact that lingers as the video ends. A faint, rhythmic electronic clicking, like a fast metronome, plays consistently in the background throughout the entire sequence, tying the dynamic cuts together.\n\nnon_diegetic_music: A fast-paced, aggressive cyberpunk-style electronic track drives the sequence, featuring a heavy, distorted 808 bassline and rapid, syncopated hi-hat trills. The music maintains a high-energy, rhythmic pulse that perfectly matches the dynamic, multi-shot fashion edit and the digital glitch aesthetic.",
142 "width": 1344,
143 "height": 768,
144 "value_1": 5,
145 "noise_seed": 793369680225544,
146 "unet_name": "fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors",
147 "clip_name": "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
148 "vae_name": "minimax_h3_video_vae_int8_convrot.safetensors",
149 "vae_name_1": "minimax_h3_audio_vae_fp32.safetensors",
150 "attention": "comfy kitchen attention"
151 }
152 },
153 {
154 "id": 136,
155 "type": "LoadImage",
156 "pos": [
157 -1230,
158 4880
159 ],
160 "size": [
161 290,
162 330
163 ],
164 "flags": [],
165 "order": 0,
166 "mode": 0,
167 "inputs": [],
168 "outputs": [
169 {
170 "name": "IMAGE",
171 "type": "IMAGE",
172 "links": [
173 262,
174 270
175 ]
176 },
177 {
178 "name": "MASK",
179 "type": "MASK",
180 "links": null
181 }
182 ],
183 "properties": {
184 "Node name for S&R": "LoadImage"
185 },
186 "widgets_values": [
187 "red_line_barrier.png",
188 "image"
189 ],
190 "widgets_values_named": {
191 "image": "red_line_barrier.png",
192 "upload": "image"
193 }
194 },
195 {
196 "id": 139,
197 "type": "MarkdownNote",
198 "pos": [
199 -1730,
200 4880
201 ],
202 "size": [
203 450,
204 630
205 ],
206 "flags": [],
207 "order": 1,
208 "mode": 0,
209 "inputs": [],
210 "outputs": [],
211 "title": "Note: Fast Video FastH3",
212 "properties": [],
213 "widgets_values": [
214 "\nThis template runs the **Image to Video** task (`MiniMaxH3ImageToVideo` node) with the [FastVideo FastH3 8-Step V2](https://huggingface.co/FastVideo/FastVideo-FastH3-8-Step-V2) checkpoint, a DMD2-distilled model that generates synchronized video and audio in **8 sampling steps** instead of the base model's full schedule. It 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**Checkpoint repo**: [FastVideo/FastVideo-FastH3-Comfy](https://huggingface.co/FastVideo/FastVideo-FastH3-Comfy)\n\n**Scope**: this distilled checkpoint supports t2va and fl2va only. Ref2VA (multi-reference conditioning) was not distilled; use the base MiniMax H3 model for reference-based tasks.\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\n"
215 ],
216 "widgets_values_named": {
217 "text": "\nThis template runs the **Image to Video** task (`MiniMaxH3ImageToVideo` node) with the [FastVideo FastH3 8-Step V2](https://huggingface.co/FastVideo/FastVideo-FastH3-8-Step-V2) checkpoint, a DMD2-distilled model that generates synchronized video and audio in **8 sampling steps** instead of the base model's full schedule. It 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**Checkpoint repo**: [FastVideo/FastVideo-FastH3-Comfy](https://huggingface.co/FastVideo/FastVideo-FastH3-Comfy)\n\n**Scope**: this distilled checkpoint supports t2va and fl2va only. Ref2VA (multi-reference conditioning) was not distilled; use the base MiniMax H3 model for reference-based tasks.\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\n"
218 },
219 "color": "#222",
220 "bgcolor": "#000"
221 },
222 {
223 "id": 140,
224 "type": "MarkdownNote",
225 "pos": [
226 -2220,
227 4880
228 ],
229 "size": [
230 440,
231 1030
232 ],
233 "flags": [],
234 "order": 2,
235 "mode": 0,
236 "inputs": [],
237 "outputs": [],
238 "title": "Note: Model Links",
239 "properties": [],
240 "widgets_values": [
241 "Guide: [Subgraph](https://docs.comfy.org/interface/features/subgraph)\n\n## Input Asset\n\n- [red_line_barrier.png](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/red_line_barrier.png)\n\n## Model Links\n\n- [Hugging Face:Comfy-Org/MiniMax-H3](https://huggingface.co/Comfy-Org/MiniMax-H3)\n\n**ModelScope mirrors** (alternative repo location)\n\n- [ModelScope:Comfy-Org/MiniMax-H3](https://modelscope.cn/models/Comfy-Org/MiniMax-H3)\n\n**vae**\n\n- [minimax_h3_video_vae_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_int8_convrot.safetensors) (2.62 GB)\n- [minimax_h3_audio_vae_fp32.safetensors](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors) (577.2 MB)\n\n**diffusion_models**\n\n- [fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors](https://huggingface.co/FastVideo/FastVideo-FastH3-Comfy/resolve/main/diffusion_models/fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors) (20.61 GB)\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) (14.61 GB)\n\n## Model Storage Location\n\n```\n📂 ComfyUI/\n├── 📂 models/\n│ ├── 📂 vae/\n│ │ ├── minimax_h3_video_vae_int8_convrot.safetensors\n│ │ └── minimax_h3_audio_vae_fp32.safetensors\n│ ├── 📂 diffusion_models/\n│ │ └── fastvideo_fasth3_8step_v2_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"
242 ],
243 "widgets_values_named": {
244 "text": "Guide: [Subgraph](https://docs.comfy.org/interface/features/subgraph)\n\n## Input Asset\n\n- [red_line_barrier.png](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/red_line_barrier.png)\n\n## Model Links\n\n- [Hugging Face:Comfy-Org/MiniMax-H3](https://huggingface.co/Comfy-Org/MiniMax-H3)\n\n**ModelScope mirrors** (alternative repo location)\n\n- [ModelScope:Comfy-Org/MiniMax-H3](https://modelscope.cn/models/Comfy-Org/MiniMax-H3)\n\n**vae**\n\n- [minimax_h3_video_vae_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_int8_convrot.safetensors) (2.62 GB)\n- [minimax_h3_audio_vae_fp32.safetensors](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors) (577.2 MB)\n\n**diffusion_models**\n\n- [fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors](https://huggingface.co/FastVideo/FastVideo-FastH3-Comfy/resolve/main/diffusion_models/fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors) (20.61 GB)\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) (14.61 GB)\n\n## Model Storage Location\n\n```\n📂 ComfyUI/\n├── 📂 models/\n│ ├── 📂 vae/\n│ │ ├── minimax_h3_video_vae_int8_convrot.safetensors\n│ │ └── minimax_h3_audio_vae_fp32.safetensors\n│ ├── 📂 diffusion_models/\n│ │ └── fastvideo_fasth3_8step_v2_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"
245 },
246 "color": "#222",
247 "bgcolor": "#000"
248 },
249 {
250 "id": 141,
251 "type": "GetImageSize",
252 "pos": [
253 -1190,
254 5480
255 ],
256 "size": [
257 230,
258 100
259 ],
260 "flags": [],
261 "order": 4,
262 "mode": 0,
263 "inputs": [
264 {
265 "name": "image",
266 "type": "IMAGE",
267 "link": 269
268 }
269 ],
270 "outputs": [
271 {
272 "name": "width",
273 "type": "INT",
274 "links": [
275 271
276 ]
277 },
278 {
279 "name": "height",
280 "type": "INT",
281 "links": [
282 272
283 ]
284 },
285 {
286 "name": "batch_size",
287 "type": "INT",
288 "links": null
289 }
290 ],
291 "properties": {
292 "Node name for S&R": "GetImageSize",
293 "cnr_id": "comfy-core",
294 "ver": "0.33.0"
295 }
296 },
297 {
298 "id": 142,
299 "type": "ImageScaleToTotalPixels",
300 "pos": [
301 -1220,
302 5270
303 ],
304 "size": [
305 280,
306 170
307 ],
308 "flags": [],
309 "order": 3,
310 "mode": 0,
311 "showAdvanced": true,
312 "inputs": [
313 {
314 "name": "image",
315 "type": "IMAGE",
316 "link": 270
317 }
318 ],
319 "outputs": [
320 {
321 "name": "IMAGE",
322 "type": "IMAGE",
323 "links": [
324 269
325 ]
326 }
327 ],
328 "properties": {
329 "Node name for S&R": "ImageScaleToTotalPixels",
330 "cnr_id": "comfy-core",
331 "ver": "0.33.0"
332 },
333 "widgets_values": [
334 "nearest-exact",
335 0.4,
336 32
337 ],
338 "widgets_values_named": {
339 "upscale_method": "nearest-exact",
340 "megapixels": 0.4,
341 "resolution_steps": 32
342 }
343 }
344 ],
345 "links": [
346 [
347 194,
348 105,
349 0,
350 92,
351 0,
352 "VIDEO"
353 ],
354 [
355 262,
356 136,
357 0,
358 105,
359 0,
360 "IMAGE"
361 ],
362 [
363 269,
364 142,
365 0,
366 141,
367 0,
368 "IMAGE"
369 ],
370 [
371 270,
372 136,
373 0,
374 142,
375 0,
376 "IMAGE"
377 ],
378 [
379 271,
380 141,
381 0,
382 105,
383 2,
384 "INT"
385 ],
386 [
387 272,
388 141,
389 1,
390 105,
391 3,
392 "INT"
393 ]
394 ],
395 "groups": [],
396 "definitions": {
397 "subgraphs": [
398 {
399 "id": "4c314f31-ecda-4b08-ae98-faaba1bf613f",
400 "version": 1,
401 "state": {
402 "lastGroupId": 7,
403 "lastNodeId": 143,
404 "lastLinkId": 274,
405 "lastRerouteId": 0
406 },
407 "revision": 0,
408 "config": [],
409 "name": "First Last Frame to Video (FastVideo FastH3)",
410 "inputNode": {
411 "id": -10,
412 "bounding": [
413 -2640,
414 4810,
415 128,
416 288
417 ]
418 },
419 "outputNode": {
420 "id": -20,
421 "bounding": [
422 2130,
423 5070,
424 128,
425 68
426 ]
427 },
428 "inputs": [
429 {
430 "id": "d6eaa195-2266-4016-b7ed-b2d17a3e53c2",
431 "name": "first_frame",
432 "type": "IMAGE",
433 "linkIds": [
434 195
435 ],
436 "pos": [
437 -2536,
438 4834
439 ]
440 },
441 {
442 "id": "03b95567-f496-4279-9d38-989dd34fa882",
443 "name": "last_frame",
444 "type": "IMAGE",
445 "linkIds": [
446 196
447 ],
448 "pos": [
449 -2536,
450 4854
451 ]
452 },
453 {
454 "id": "d7302ca7-24ed-44c4-8ac9-736540dab7fb",
455 "name": "prompt",
456 "type": "STRING",
457 "linkIds": [
458 197
459 ],
460 "pos": [
461 -2536,
462 4874
463 ]
464 },
465 {
466 "id": "709e2d94-3172-496f-ad86-7d672f3df568",
467 "name": "width",
468 "type": "INT",
469 "linkIds": [
470 200
471 ],
472 "pos": [
473 -2536,
474 4894
475 ]
476 },
477 {
478 "id": "8205b85b-19bb-47f6-8418-89cd42940c1d",
479 "name": "height",
480 "type": "INT",
481 "linkIds": [
482 201
483 ],
484 "pos": [
485 -2536,
486 4914
487 ]
488 },
489 {
490 "id": "a40e5e96-4307-4a8e-a4f8-13a27d4bc9d3",
491 "name": "value_1",
492 "type": "FLOAT",
493 "linkIds": [
494 206
495 ],
496 "label": "duration",
497 "pos": [
498 -2536,
499 4934
500 ]
501 },
502 {
503 "id": "9f8734bc-c2b3-43b0-911e-e89560e8b777",
504 "name": "noise_seed",
505 "type": "INT",
506 "linkIds": [
507 207
508 ],
509 "pos": [
510 -2536,
511 4954
512 ]
513 },
514 {
515 "id": "9a6d2811-673c-47e0-aa94-dda7c83621e4",
516 "name": "unet_name",
517 "type": "COMBO",
518 "linkIds": [
519 221
520 ],
521 "pos": [
522 -2536,
523 4974
524 ]
525 },
526 {
527 "id": "7100a787-5536-4002-840b-1d2fd071010a",
528 "name": "clip_name",
529 "type": "COMBO",
530 "linkIds": [
531 227
532 ],
533 "pos": [
534 -2536,
535 4994
536 ]
537 },
538 {
539 "id": "d40632fd-e0ae-4319-b85d-5479db4d6e11",
540 "name": "vae_name",
541 "type": "COMBO",
542 "linkIds": [
543 223
544 ],
545 "pos": [
546 -2536,
547 5014
548 ]
549 },
550 {
551 "id": "2992852c-4b20-439a-8771-c866ec1996e6",
552 "name": "vae_name_1",
553 "type": "COMBO",
554 "linkIds": [
555 224
556 ],
557 "label": "audio_vae",
558 "pos": [
559 -2536,
560 5034
561 ]
562 },
563 {
564 "id": "42566ff0-1439-4215-9724-33d6c0282e21",
565 "name": "attention",
566 "type": "COMBO",
567 "linkIds": [
568 258
569 ],
570 "pos": [
571 -2536,
572 5054
573 ]
574 }
575 ],
576 "outputs": [
577 {
578 "id": "adb2611c-e490-4fca-8067-b718c992f8cc",
579 "name": "VIDEO",
580 "type": "VIDEO",
581 "linkIds": [
582 168
583 ],
584 "localized_name": "VIDEO",
585 "pos": [
586 2154,
587 5094
588 ]
589 }
590 ],
591 "widgets": [],
592 "nodes": [
593 {
594 "id": 11,
595 "type": "VAELoader",
596 "pos": [
597 -2030,
598 5110
599 ],
600 "size": [
601 640,
602 100
603 ],
604 "flags": [],
605 "order": 4,
606 "mode": 0,
607 "inputs": [
608 {
609 "localized_name": "vae_name",
610 "name": "vae_name",
611 "type": "COMBO",
612 "widget": {
613 "name": "vae_name"
614 },
615 "link": 223
616 }
617 ],
618 "outputs": [
619 {
620 "localized_name": "VAE",
621 "name": "VAE",
622 "type": "VAE",
623 "links": [
624 8,
625 190
626 ]
627 }
628 ],
629 "properties": {
630 "Node name for S&R": "VAELoader",
631 "cnr_id": "comfy-core",
632 "ver": "0.33.0",
633 "models": [
634 {
635 "name": "minimax_h3_video_vae_int8_convrot.safetensors",
636 "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_int8_convrot.safetensors",
637 "directory": "vae"
638 }
639 ]
640 },
641 "widgets_values": [
642 "minimax_h3_video_vae_int8_convrot.safetensors"
643 ],
644 "widgets_values_named": {
645 "vae_name": "minimax_h3_video_vae_int8_convrot.safetensors"
646 }
647 },
648 {
649 "id": 24,
650 "type": "VAELoader",
651 "pos": [
652 -2030,
653 5290
654 ],
655 "size": [
656 650,
657 100
658 ],
659 "flags": [],
660 "order": 10,
661 "mode": 0,
662 "inputs": [
663 {
664 "localized_name": "vae_name",
665 "name": "vae_name",
666 "type": "COMBO",
667 "widget": {
668 "name": "vae_name"
669 },
670 "link": 224
671 }
672 ],
673 "outputs": [
674 {
675 "localized_name": "VAE",
676 "name": "VAE",
677 "type": "VAE",
678 "links": [
679 23
680 ]
681 }
682 ],
683 "properties": {
684 "Node name for S&R": "VAELoader",
685 "cnr_id": "comfy-core",
686 "ver": "0.33.0",
687 "models": [
688 {
689 "name": "minimax_h3_audio_vae_fp32.safetensors",
690 "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors",
691 "directory": "vae"
692 }
693 ]
694 },
695 "widgets_values": [
696 "minimax_h3_audio_vae_fp32.safetensors"
697 ],
698 "widgets_values_named": {
699 "vae_name": "minimax_h3_audio_vae_fp32.safetensors"
700 }
701 },
702 {
703 "id": 23,
704 "type": "VAEDecodeAudio",
705 "pos": [
706 10,
707 4990
708 ],
709 "size": [
710 230,
711 80
712 ],
713 "flags": {
714 "collapsed": false
715 },
716 "order": 9,
717 "mode": 0,
718 "inputs": [
719 {
720 "localized_name": "samples",
721 "name": "samples",
722 "type": "LATENT",
723 "link": 226
724 },
725 {
726 "localized_name": "vae",
727 "name": "vae",
728 "type": "VAE",
729 "link": 23
730 }
731 ],
732 "outputs": [
733 {
734 "localized_name": "AUDIO",
735 "name": "AUDIO",
736 "type": "AUDIO",
737 "links": [
738 166
739 ]
740 }
741 ],
742 "properties": {
743 "Node name for S&R": "VAEDecodeAudio",
744 "cnr_id": "comfy-core",
745 "ver": "0.33.0"
746 }
747 },
748 {
749 "id": 10,
750 "type": "VAEDecode",
751 "pos": [
752 10,
753 4850
754 ],
755 "size": [
756 230,
757 80
758 ],
759 "flags": {
760 "collapsed": false
761 },
762 "order": 3,
763 "mode": 0,
764 "inputs": [
765 {
766 "localized_name": "samples",
767 "name": "samples",
768 "type": "LATENT",
769 "link": 225
770 },
771 {
772 "localized_name": "vae",
773 "name": "vae",
774 "type": "VAE",
775 "link": 8
776 }
777 ],
778 "outputs": [
779 {
780 "localized_name": "IMAGE",
781 "name": "IMAGE",
782 "type": "IMAGE",
783 "links": [
784 167
785 ]
786 }
787 ],
788 "properties": {
789 "Node name for S&R": "VAEDecode",
790 "cnr_id": "comfy-core",
791 "ver": "0.33.0"
792 }
793 },
794 {
795 "id": 17,
796 "type": "KSamplerSelect",
797 "pos": [
798 -760,
799 4950
800 ],
801 "size": [
802 370,
803 90
804 ],
805 "flags": [],
806 "order": 0,
807 "mode": 0,
808 "inputs": [],
809 "outputs": [
810 {
811 "localized_name": "SAMPLER",
812 "name": "SAMPLER",
813 "type": "SAMPLER",
814 "links": [
815 16
816 ]
817 }
818 ],
819 "properties": {
820 "Node name for S&R": "KSamplerSelect",
821 "cnr_id": "comfy-core",
822 "ver": "0.33.0"
823 },
824 "widgets_values": [
825 "res_multistep"
826 ],
827 "widgets_values_named": {
828 "sampler_name": "res_multistep"
829 }
830 },
831 {
832 "id": 9,
833 "type": "BasicScheduler",
834 "pos": [
835 -760,
836 5100
837 ],
838 "size": [
839 370,
840 150
841 ],
842 "flags": [],
843 "order": 2,
844 "mode": 0,
845 "inputs": [
846 {
847 "localized_name": "model",
848 "name": "model",
849 "type": "MODEL",
850 "link": 266
851 }
852 ],
853 "outputs": [
854 {
855 "localized_name": "SIGMAS",
856 "name": "SIGMAS",
857 "type": "SIGMAS",
858 "links": [
859 18
860 ]
861 }
862 ],
863 "properties": {
864 "Node name for S&R": "BasicScheduler",
865 "cnr_id": "comfy-core",
866 "ver": "0.33.0"
867 },
868 "widgets_values": [
869 "simple",
870 8,
871 1
872 ],
873 "widgets_values_named": {
874 "scheduler": "simple",
875 "steps": 8,
876 "denoise": 1
877 }
878 },
879 {
880 "id": 14,
881 "type": "SamplerCustomAdvanced",
882 "pos": [
883 -340,
884 4780
885 ],
886 "size": [
887 230,
888 390
889 ],
890 "flags": [],
891 "order": 6,
892 "mode": 0,
893 "inputs": [
894 {
895 "localized_name": "noise",
896 "name": "noise",
897 "type": "NOISE",
898 "link": 40
899 },
900 {
901 "localized_name": "guider",
902 "name": "guider",
903 "type": "GUIDER",
904 "link": 12
905 },
906 {
907 "localized_name": "sampler",
908 "name": "sampler",
909 "type": "SAMPLER",
910 "link": 16
911 },
912 {
913 "localized_name": "sigmas",
914 "name": "sigmas",
915 "type": "SIGMAS",
916 "link": 18
917 },
918 {
919 "localized_name": "latent_image",
920 "name": "latent_image",
921 "type": "LATENT",
922 "link": 188
923 }
924 ],
925 "outputs": [
926 {
927 "localized_name": "output",
928 "name": "output",
929 "type": "LATENT",
930 "links": [
931 225,
932 226
933 ]
934 },
935 {
936 "localized_name": "denoised_output",
937 "name": "denoised_output",
938 "type": "LATENT",
939 "links": null
940 }
941 ],
942 "properties": {
943 "Node name for S&R": "SamplerCustomAdvanced",
944 "cnr_id": "comfy-core",
945 "ver": "0.33.0"
946 }
947 },
948 {
949 "id": 16,
950 "type": "BasicGuider",
951 "pos": [
952 -760,
953 4820
954 ],
955 "size": [
956 360,
957 80
958 ],
959 "flags": [],
960 "order": 8,
961 "mode": 0,
962 "inputs": [
963 {
964 "localized_name": "model",
965 "name": "model",
966 "type": "MODEL",
967 "link": 267
968 },
969 {
970 "localized_name": "conditioning",
971 "name": "conditioning",
972 "type": "CONDITIONING",
973 "link": 187
974 }
975 ],
976 "outputs": [
977 {
978 "localized_name": "GUIDER",
979 "name": "GUIDER",
980 "type": "GUIDER",
981 "links": [
982 12
983 ]
984 }
985 ],
986 "properties": {
987 "Node name for S&R": "BasicGuider",
988 "cnr_id": "comfy-core",
989 "ver": "0.33.0"
990 }
991 },
992 {
993 "id": 6,
994 "type": "UNETLoader",
995 "pos": [
996 -2030,
997 4690
998 ],
999 "size": [
1000 640,
1001 120
1002 ],
1003 "flags": [],
1004 "order": 1,
1005 "mode": 0,
1006 "inputs": [
1007 {
1008 "localized_name": "unet_name",
1009 "name": "unet_name",
1010 "type": "COMBO",
1011 "widget": {
1012 "name": "unet_name"
1013 },
1014 "link": 221
1015 }
1016 ],
1017 "outputs": [
1018 {
1019 "localized_name": "MODEL",
1020 "name": "MODEL",
1021 "type": "MODEL",
1022 "links": [
1023 274
1024 ]
1025 }
1026 ],
1027 "properties": {
1028 "Node name for S&R": "UNETLoader",
1029 "cnr_id": "comfy-core",
1030 "ver": "0.33.0",
1031 "models": [
1032 {
1033 "name": "fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors",
1034 "url": "https://huggingface.co/FastVideo/FastVideo-FastH3-Comfy/resolve/main/diffusion_models/fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors",
1035 "directory": "diffusion_models"
1036 }
1037 ]
1038 },
1039 "widgets_values": [
1040 "fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors",
1041 "default"
1042 ],
1043 "widgets_values_named": {
1044 "unet_name": "fastvideo_fasth3_8step_v2_pruned_int8_convrot.safetensors",
1045 "weight_dtype": "default"
1046 }
1047 },
1048 {
1049 "id": 13,
1050 "type": "CLIPLoader",
1051 "pos": [
1052 -2030,
1053 4880
1054 ],
1055 "size": [
1056 640,
1057 150
1058 ],
1059 "flags": [],
1060 "order": 5,
1061 "mode": 0,
1062 "inputs": [
1063 {
1064 "localized_name": "clip_name",
1065 "name": "clip_name",
1066 "type": "COMBO",
1067 "widget": {
1068 "name": "clip_name"
1069 },
1070 "link": 227
1071 }
1072 ],
1073 "outputs": [
1074 {
1075 "localized_name": "CLIP",
1076 "name": "CLIP",
1077 "type": "CLIP",
1078 "links": [
1079 189
1080 ]
1081 }
1082 ],
1083 "properties": {
1084 "Node name for S&R": "CLIPLoader",
1085 "cnr_id": "comfy-core",
1086 "ver": "0.33.0",
1087 "models": [
1088 {
1089 "name": "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
1090 "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
1091 "directory": "text_encoders"
1092 }
1093 ]
1094 },
1095 "widgets_values": [
1096 "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
1097 "minimax",
1098 "default"
1099 ],
1100 "widgets_values_named": {
1101 "clip_name": "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
1102 "type": "minimax",
1103 "device": "default"
1104 }
1105 },
1106 {
1107 "id": 15,
1108 "type": "RandomNoise",
1109 "pos": [
1110 -760,
1111 4660
1112 ],
1113 "size": [
1114 360,
1115 90
1116 ],
1117 "flags": [],
1118 "order": 7,
1119 "mode": 0,
1120 "inputs": [
1121 {
1122 "localized_name": "noise_seed",
1123 "name": "noise_seed",
1124 "type": "INT",
1125 "widget": {
1126 "name": "noise_seed"
1127 },
1128 "link": 207
1129 }
1130 ],
1131 "outputs": [
1132 {
1133 "localized_name": "NOISE",
1134 "name": "NOISE",
1135 "type": "NOISE",
1136 "links": [
1137 40
1138 ]
1139 }
1140 ],
1141 "properties": {
1142 "Node name for S&R": "RandomNoise",
1143 "cnr_id": "comfy-core",
1144 "ver": "0.33.0"
1145 },
1146 "widgets_values": [
1147 1,
1148 "fixed"
1149 ],
1150 "widgets_values_named": {
1151 "noise_seed": 1,
1152 "control_after_generate": "fixed"
1153 }
1154 },
1155 {
1156 "id": 91,
1157 "type": "CreateVideo",
1158 "pos": [
1159 340,
1160 4890
1161 ],
1162 "size": [
1163 270,
1164 170
1165 ],
1166 "flags": [],
1167 "order": 11,
1168 "mode": 0,
1169 "inputs": [
1170 {
1171 "localized_name": "images",
1172 "name": "images",
1173 "type": "IMAGE",
1174 "link": 167
1175 },
1176 {
1177 "localized_name": "audio",
1178 "name": "audio",
1179 "shape": 7,
1180 "type": "AUDIO",
1181 "link": 166
1182 }
1183 ],
1184 "outputs": [
1185 {
1186 "localized_name": "VIDEO",
1187 "name": "VIDEO",
1188 "type": "VIDEO",
1189 "links": [
1190 168
1191 ]
1192 }
1193 ],
1194 "properties": {
1195 "Node name for S&R": "CreateVideo",
1196 "cnr_id": "comfy-core",
1197 "ver": "0.33.0"
1198 },
1199 "widgets_values": [
1200 24,
1201 "auto",
1202 "sRGB",
1203 "none"
1204 ],
1205 "widgets_values_named": {
1206 "fps": 24,
1207 "bit_depth": "auto",
1208 "color_space": "sRGB",
1209 "codec": "none"
1210 }
1211 },
1212 {
1213 "id": 104,
1214 "type": "MiniMaxH3ImageToVideo",
1215 "pos": [
1216 -1290,
1217 4740
1218 ],
1219 "size": [
1220 430,
1221 550
1222 ],
1223 "flags": [],
1224 "order": 12,
1225 "mode": 0,
1226 "inputs": [
1227 {
1228 "localized_name": "clip",
1229 "name": "clip",
1230 "type": "CLIP",
1231 "link": 189
1232 },
1233 {
1234 "localized_name": "vae",
1235 "name": "vae",
1236 "type": "VAE",
1237 "link": 190
1238 },
1239 {
1240 "localized_name": "first_frame",
1241 "name": "first_frame",
1242 "shape": 7,
1243 "type": "IMAGE",
1244 "link": 195
1245 },
1246 {
1247 "localized_name": "last_frame",
1248 "name": "last_frame",
1249 "shape": 7,
1250 "type": "IMAGE",
1251 "link": 196
1252 },
1253 {
1254 "localized_name": "prompt",
1255 "name": "prompt",
1256 "type": "STRING",
1257 "widget": {
1258 "name": "prompt"
1259 },
1260 "link": 197
1261 },
1262 {
1263 "localized_name": "width",
1264 "name": "width",
1265 "type": "INT",
1266 "widget": {
1267 "name": "width"
1268 },
1269 "link": 200
1270 },
1271 {
1272 "localized_name": "height",
1273 "name": "height",
1274 "type": "INT",
1275 "widget": {
1276 "name": "height"
1277 },
1278 "link": 201
1279 },
1280 {
1281 "localized_name": "length",
1282 "name": "length",
1283 "type": "INT",
1284 "widget": {
1285 "name": "length"
1286 },
1287 "link": 199
1288 }
1289 ],
1290 "outputs": [
1291 {
1292 "localized_name": "positive",
1293 "name": "positive",
1294 "type": "CONDITIONING",
1295 "links": [
1296 187
1297 ]
1298 },
1299 {
1300 "localized_name": "LATENT",
1301 "name": "LATENT",
1302 "type": "LATENT",
1303 "links": [
1304 188
1305 ]
1306 }
1307 ],
1308 "properties": {
1309 "Node name for S&R": "MiniMaxH3ImageToVideo",
1310 "cnr_id": "comfy-core",
1311 "ver": "0.33.0"
1312 },
1313 "widgets_values": [
1314 "Vaporwave title sequence look: pink and blue gradient palette, VHS tracking artifacts, Greek statue motifs, chrome palm trees, RGB chromatic aberration, lo-fi retro atmosphere, mood languid and nostalgic.\n\nTimeline:\n[0s-1s] VHS static opens the frame, the title \"COMFYUI\" appears with RGB split and a slight horizontal jitter.\n[1s-2.5s] Hard cut, a Greek plaster bust close-up, pink-purple gradient sky, a pixelated sun.\n[2.5s-4s] Clean \"STARRING\" credits appear, \"LATENT\" and \"CONTROLNET\" each shown exactly once.\n[4s-5s] Final card \"DIRECTED BY COMFYUI\" holds, one VHS tracking glitch settling into stability.\n\nHard cuts only, transitions landing with tape jumps, no push-ins, no dissolves.\n\nAudio: lo-fi vaporwave score, slow drum machine with soft bass, VHS tape-noise sample joins at 2.5s, melody fading for the last 1s.\n\nAll text must be clearly legible, do not misspell English, no Chinese characters, do not repeat names or job titles, no soft dissolves, no subtitle bars.",
1315 1344,
1316 768,
1317 73
1318 ],
1319 "widgets_values_named": {
1320 "prompt": "Vaporwave title sequence look: pink and blue gradient palette, VHS tracking artifacts, Greek statue motifs, chrome palm trees, RGB chromatic aberration, lo-fi retro atmosphere, mood languid and nostalgic.\n\nTimeline:\n[0s-1s] VHS static opens the frame, the title \"COMFYUI\" appears with RGB split and a slight horizontal jitter.\n[1s-2.5s] Hard cut, a Greek plaster bust close-up, pink-purple gradient sky, a pixelated sun.\n[2.5s-4s] Clean \"STARRING\" credits appear, \"LATENT\" and \"CONTROLNET\" each shown exactly once.\n[4s-5s] Final card \"DIRECTED BY COMFYUI\" holds, one VHS tracking glitch settling into stability.\n\nHard cuts only, transitions landing with tape jumps, no push-ins, no dissolves.\n\nAudio: lo-fi vaporwave score, slow drum machine with soft bass, VHS tape-noise sample joins at 2.5s, melody fading for the last 1s.\n\nAll text must be clearly legible, do not misspell English, no Chinese characters, do not repeat names or job titles, no soft dissolves, no subtitle bars.",
1321 "width": 1344,
1322 "height": 768,
1323 "length": 73
1324 }
1325 },
1326 {
1327 "id": 107,
1328 "type": "ComfyMathExpression",
1329 "pos": [
1330 -1720,
1331 5600
1332 ],
1333 "size": [
1334 360,
1335 170
1336 ],
1337 "flags": {
1338 "collapsed": false
1339 },
1340 "order": 13,
1341 "mode": 0,
1342 "inputs": [
1343 {
1344 "label": "a",
1345 "localized_name": "a",
1346 "name": "values.a",
1347 "type": "FLOAT,INT,BOOLEAN",
1348 "link": 205
1349 },
1350 {
1351 "localized_name": "b",
1352 "name": "values.b",
1353 "shape": 7,
1354 "type": "FLOAT,INT,BOOLEAN",
1355 "link": null
1356 }
1357 ],
1358 "outputs": [
1359 {
1360 "localized_name": "FLOAT",
1361 "name": "FLOAT",
1362 "type": "FLOAT",
1363 "links": null
1364 },
1365 {
1366 "localized_name": "INT",
1367 "name": "INT",
1368 "type": "INT",
1369 "links": [
1370 199
1371 ]
1372 },
1373 {
1374 "localized_name": "BOOL",
1375 "name": "BOOL",
1376 "type": "BOOLEAN",
1377 "links": null
1378 }
1379 ],
1380 "properties": {
1381 "Node name for S&R": "ComfyMathExpression",
1382 "cnr_id": "comfy-core",
1383 "ver": "0.33.0"
1384 },
1385 "widgets_values": [
1386 "max(5, round(a * 24)) + (5 - (max(5, round(a * 24)) % 17)) % 17"
1387 ],
1388 "widgets_values_named": {
1389 "expression": "max(5, round(a * 24)) + (5 - (max(5, round(a * 24)) % 17)) % 17"
1390 }
1391 },
1392 {
1393 "id": 111,
1394 "type": "PrimitiveFloat",
1395 "pos": [
1396 -2030,
1397 5600
1398 ],
1399 "size": [
1400 270,
1401 90
1402 ],
1403 "flags": [],
1404 "order": 14,
1405 "mode": 0,
1406 "inputs": [
1407 {
1408 "localized_name": "value",
1409 "name": "value",
1410 "type": "FLOAT",
1411 "widget": {
1412 "name": "value"
1413 },
1414 "link": 206
1415 }
1416 ],
1417 "outputs": [
1418 {
1419 "localized_name": "FLOAT",
1420 "name": "FLOAT",
1421 "type": "FLOAT",
1422 "links": [
1423 205
1424 ]
1425 }
1426 ],
1427 "title": "Float (duration)",
1428 "properties": {
1429 "Node name for S&R": "PrimitiveFloat",
1430 "cnr_id": "comfy-core",
1431 "ver": "0.33.0"
1432 },
1433 "widgets_values": [
1434 2
1435 ],
1436 "widgets_values_named": {
1437 "value": 2
1438 }
1439 },
1440 {
1441 "id": 127,
1442 "type": "BlockSparseAttention",
1443 "pos": [
1444 -660,
1445 4200
1446 ],
1447 "size": [
1448 270,
1449 250
1450 ],
1451 "flags": [],
1452 "order": 15,
1453 "mode": 0,
1454 "inputs": [
1455 {
1456 "localized_name": "model",
1457 "name": "model",
1458 "type": "MODEL",
1459 "link": 249
1460 }
1461 ],
1462 "outputs": [
1463 {
1464 "localized_name": "model",
1465 "name": "model",
1466 "type": "MODEL",
1467 "links": [
1468 266,
1469 267
1470 ]
1471 }
1472 ],
1473 "properties": {
1474 "Node name for S&R": "BlockSparseAttention"
1475 },
1476 "widgets_values": [
1477 "vsa",
1478 10,
1479 0.2,
1480 1,
1481 "",
1482 12288,
1483 256,
1484 "exact_kv_and_rows",
1485 false
1486 ],
1487 "widgets_values_named": {
1488 "selection": "vsa",
1489 "selection.keep_percent": 10,
1490 "start_percent": 0.2,
1491 "end_percent": 1,
1492 "dense_blocks": "",
1493 "min_tokens": 12288,
1494 "extra_tokens": 256,
1495 "sink_conditioning": "exact_kv_and_rows",
1496 "verbose": false
1497 }
1498 },
1499 {
1500 "id": 128,
1501 "type": "ModelAttentionBackend",
1502 "pos": [
1503 -1010,
1504 4200
1505 ],
1506 "size": [
1507 300,
1508 130
1509 ],
1510 "flags": [],
1511 "order": 16,
1512 "mode": 0,
1513 "inputs": [
1514 {
1515 "localized_name": "model",
1516 "name": "model",
1517 "type": "MODEL",
1518 "link": 273
1519 },
1520 {
1521 "localized_name": "backend",
1522 "name": "attention",
1523 "type": "COMBO",
1524 "widget": {
1525 "name": "attention"
1526 },
1527 "link": 258
1528 }
1529 ],
1530 "outputs": [
1531 {
1532 "localized_name": "model",
1533 "name": "model",
1534 "type": "MODEL",
1535 "links": [
1536 249
1537 ]
1538 }
1539 ],
1540 "properties": {
1541 "Node name for S&R": "ModelAttentionBackend"
1542 },
1543 "widgets_values": [
1544 "comfy kitchen attention"
1545 ],
1546 "widgets_values_named": {
1547 "attention": "comfy kitchen attention"
1548 }
1549 },
1550 {
1551 "id": 143,
1552 "type": "MiniMaxH3SigmaShift",
1553 "pos": [
1554 -1640,
1555 4210
1556 ],
1557 "size": [
1558 280,
1559 90
1560 ],
1561 "flags": [],
1562 "order": 17,
1563 "mode": 0,
1564 "inputs": [
1565 {
1566 "localized_name": "model",
1567 "name": "model",
1568 "type": "MODEL",
1569 "link": 274
1570 }
1571 ],
1572 "outputs": [
1573 {
1574 "localized_name": "MODEL",
1575 "name": "MODEL",
1576 "type": "MODEL",
1577 "links": [
1578 273
1579 ]
1580 }
1581 ],
1582 "properties": {
1583 "Node name for S&R": "MiniMaxH3SigmaShift"
1584 },
1585 "widgets_values": [
1586 10,
1587 3
1588 ],
1589 "widgets_values_named": {
1590 "shift_video": 10,
1591 "shift_audio": 3
1592 }
1593 }
1594 ],
1595 "groups": [
1596 {
1597 "id": 1,
1598 "title": "Models",
1599 "bounding": [
1600 -2050,
1601 4540,
1602 700,
1603 980
1604 ],
1605 "color": "#3f789e",
1606 "flags": []
1607 },
1608 {
1609 "id": 2,
1610 "title": "Sampling",
1611 "bounding": [
1612 -790,
1613 4540,
1614 700,
1615 980
1616 ],
1617 "color": "#3f789e",
1618 "flags": []
1619 },
1620 {
1621 "id": 3,
1622 "title": "Conditioning",
1623 "bounding": [
1624 -1310,
1625 4540,
1626 480,
1627 980
1628 ],
1629 "color": "#3f789e",
1630 "flags": []
1631 },
1632 {
1633 "id": 4,
1634 "title": "Decoding and create video",
1635 "bounding": [
1636 -50,
1637 4540,
1638 700,
1639 980
1640 ],
1641 "color": "#3f789e",
1642 "flags": []
1643 },
1644 {
1645 "id": 7,
1646 "title": "Sparse Attention (VSA)",
1647 "bounding": [
1648 -1310,
1649 4100,
1650 1220,
1651 420
1652 ],
1653 "color": "#3f789e",
1654 "flags": []
1655 }
1656 ],
1657 "links": [
1658 {
1659 "id": 23,
1660 "origin_id": 24,
1661 "origin_slot": 0,
1662 "target_id": 23,
1663 "target_slot": 1,
1664 "type": "VAE"
1665 },
1666 {
1667 "id": 8,
1668 "origin_id": 11,
1669 "origin_slot": 0,
1670 "target_id": 10,
1671 "target_slot": 1,
1672 "type": "VAE"
1673 },
1674 {
1675 "id": 40,
1676 "origin_id": 15,
1677 "origin_slot": 0,
1678 "target_id": 14,
1679 "target_slot": 0,
1680 "type": "NOISE"
1681 },
1682 {
1683 "id": 12,
1684 "origin_id": 16,
1685 "origin_slot": 0,
1686 "target_id": 14,
1687 "target_slot": 1,
1688 "type": "GUIDER"
1689 },
1690 {
1691 "id": 16,
1692 "origin_id": 17,
1693 "origin_slot": 0,
1694 "target_id": 14,
1695 "target_slot": 2,
1696 "type": "SAMPLER"
1697 },
1698 {
1699 "id": 18,
1700 "origin_id": 9,
1701 "origin_slot": 0,
1702 "target_id": 14,
1703 "target_slot": 3,
1704 "type": "SIGMAS"
1705 },
1706 {
1707 "id": 188,
1708 "origin_id": 104,
1709 "origin_slot": 1,
1710 "target_id": 14,
1711 "target_slot": 4,
1712 "type": "LATENT"
1713 },
1714 {
1715 "id": 187,
1716 "origin_id": 104,
1717 "origin_slot": 0,
1718 "target_id": 16,
1719 "target_slot": 1,
1720 "type": "CONDITIONING"
1721 },
1722 {
1723 "id": 167,
1724 "origin_id": 10,
1725 "origin_slot": 0,
1726 "target_id": 91,
1727 "target_slot": 0,
1728 "type": "IMAGE"
1729 },
1730 {
1731 "id": 166,
1732 "origin_id": 23,
1733 "origin_slot": 0,
1734 "target_id": 91,
1735 "target_slot": 1,
1736 "type": "AUDIO"
1737 },
1738 {
1739 "id": 189,
1740 "origin_id": 13,
1741 "origin_slot": 0,
1742 "target_id": 104,
1743 "target_slot": 0,
1744 "type": "CLIP"
1745 },
1746 {
1747 "id": 190,
1748 "origin_id": 11,
1749 "origin_slot": 0,
1750 "target_id": 104,
1751 "target_slot": 1,
1752 "type": "VAE"
1753 },
1754 {
1755 "id": 168,
1756 "origin_id": 91,
1757 "origin_slot": 0,
1758 "target_id": -20,
1759 "target_slot": 0,
1760 "type": "VIDEO"
1761 },
1762 {
1763 "id": 195,
1764 "origin_id": -10,
1765 "origin_slot": 0,
1766 "target_id": 104,
1767 "target_slot": 2,
1768 "type": "IMAGE"
1769 },
1770 {
1771 "id": 196,
1772 "origin_id": -10,
1773 "origin_slot": 1,
1774 "target_id": 104,
1775 "target_slot": 3,
1776 "type": "IMAGE"
1777 },
1778 {
1779 "id": 197,
1780 "origin_id": -10,
1781 "origin_slot": 2,
1782 "target_id": 104,
1783 "target_slot": 4,
1784 "type": "STRING"
1785 },
1786 {
1787 "id": 199,
1788 "origin_id": 107,
1789 "origin_slot": 1,
1790 "target_id": 104,
1791 "target_slot": 7,
1792 "type": "INT"
1793 },
1794 {
1795 "id": 200,
1796 "origin_id": -10,
1797 "origin_slot": 3,
1798 "target_id": 104,
1799 "target_slot": 5,
1800 "type": "INT"
1801 },
1802 {
1803 "id": 201,
1804 "origin_id": -10,
1805 "origin_slot": 4,
1806 "target_id": 104,
1807 "target_slot": 6,
1808 "type": "INT"
1809 },
1810 {
1811 "id": 205,
1812 "origin_id": 111,
1813 "origin_slot": 0,
1814 "target_id": 107,
1815 "target_slot": 0,
1816 "type": "FLOAT"
1817 },
1818 {
1819 "id": 206,
1820 "origin_id": -10,
1821 "origin_slot": 5,
1822 "target_id": 111,
1823 "target_slot": 0,
1824 "type": "FLOAT"
1825 },
1826 {
1827 "id": 207,
1828 "origin_id": -10,
1829 "origin_slot": 6,
1830 "target_id": 15,
1831 "target_slot": 0,
1832 "type": "INT"
1833 },
1834 {
1835 "id": 221,
1836 "origin_id": -10,
1837 "origin_slot": 7,
1838 "target_id": 6,
1839 "target_slot": 0,
1840 "type": "COMBO"
1841 },
1842 {
1843 "id": 223,
1844 "origin_id": -10,
1845 "origin_slot": 9,
1846 "target_id": 11,
1847 "target_slot": 0,
1848 "type": "COMBO"
1849 },
1850 {
1851 "id": 224,
1852 "origin_id": -10,
1853 "origin_slot": 10,
1854 "target_id": 24,
1855 "target_slot": 0,
1856 "type": "COMBO"
1857 },
1858 {
1859 "id": 225,
1860 "origin_id": 14,
1861 "origin_slot": 0,
1862 "target_id": 10,
1863 "target_slot": 0,
1864 "type": "LATENT"
1865 },
1866 {
1867 "id": 226,
1868 "origin_id": 14,
1869 "origin_slot": 0,
1870 "target_id": 23,
1871 "target_slot": 0,
1872 "type": "LATENT"
1873 },
1874 {
1875 "id": 227,
1876 "origin_id": -10,
1877 "origin_slot": 8,
1878 "target_id": 13,
1879 "target_slot": 0,
1880 "type": "COMBO"
1881 },
1882 {
1883 "id": 249,
1884 "origin_id": 128,
1885 "origin_slot": 0,
1886 "target_id": 127,
1887 "target_slot": 0,
1888 "type": "MODEL"
1889 },
1890 {
1891 "id": 258,
1892 "origin_id": -10,
1893 "origin_slot": 11,
1894 "target_id": 128,
1895 "target_slot": 1,
1896 "type": "COMBO"
1897 },
1898 {
1899 "id": 266,
1900 "origin_id": 127,
1901 "origin_slot": 0,
1902 "target_id": 9,
1903 "target_slot": 0,
1904 "type": "MODEL"
1905 },
1906 {
1907 "id": 267,
1908 "origin_id": 127,
1909 "origin_slot": 0,
1910 "target_id": 16,
1911 "target_slot": 0,
1912 "type": "MODEL"
1913 },
1914 {
1915 "id": 273,
1916 "origin_id": 143,
1917 "origin_slot": 0,
1918 "target_id": 128,
1919 "target_slot": 0,
1920 "type": "MODEL"
1921 },
1922 {
1923 "id": 274,
1924 "origin_id": 6,
1925 "origin_slot": 0,
1926 "target_id": 143,
1927 "target_slot": 0,
1928 "type": "MODEL"
1929 }
1930 ],
1931 "extra": []
1932 }
1933 ]
1934 },
1935 "config": [],
1936 "extra": {
1937 "ds": {
1938 "scale": 0.9557986589953105,
1939 "offset": [
1940 2551.813555810394,
1941 -4988.081568095423
1942 ]
1943 },
1944 "frontendVersion": "1.52.7",
1945 "VHS_KeepIntermediate": true,
1946 "VHS_MetadataImage": true,
1947 "VHS_latentpreview": false,
1948 "VHS_latentpreviewrate": 0
1949 },
1950 "version": 0.4
1951}