stablediffusiontutorials/Minimax-H3

1{
2  "1": {
3    "inputs": {
4      "unet_name": "minimax_h3_ref2va_pruned_int8_convrot.safetensors",
5      "weight_dtype": "default"
6    },
7    "class_type": "UNETLoader",
8    "_meta": {
9      "title": "拡散モデルを読み込む"
10    }
11  },
12  "5": {
13    "inputs": {
14      "lora_name": "minimax-h3\\minimax_h3_ref2v_turbo_4step_v0.1_comfyui_bf16.safetensors",
15      "strength_model": 1,
16      "model": [
17        "1",
18        0
19      ]
20    },
21    "class_type": "LoraLoaderModelOnly",
22    "_meta": {
23      "title": "LoRAローダーモデルのみ"
24    }
25  },
26  "8": {
27    "inputs": {
28      "shift_video": 12,
29      "shift_audio": 3,
30      "model": [
31        "5",
32        0
33      ]
34    },
35    "class_type": "MiniMaxH3SigmaShift",
36    "_meta": {
37      "title": "ModelSamplingMiniMaxH3"
38    }
39  },
40  "10": {
41    "inputs": {
42      "clip_name": "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
43      "type": "minimax",
44      "device": "default"
45    },
46    "class_type": "CLIPLoader",
47    "_meta": {
48      "title": "CLIPを読み込む"
49    }
50  },
51  "11": {
52    "inputs": {
53      "image": "testS2_T03_ref_01_full_528x768.webp"
54    },
55    "class_type": "LoadImage",
56    "_meta": {
57      "title": "画像を読み込む"
58    }
59  },
60  "12": {
61    "inputs": {
62      "prompt": "subject_definitions:\n<Subject 1> is the same young anime-style girl whose identity, face, hairstyle, eye shape, and body proportions are jointly defined by <Picture 1>, <Picture 2>, and <Picture 3>, and whose dance motion comes from <Video 1>.\n<Subject 2> is a Japanese summer school uniform consisting of a short-sleeved white sailor shirt with a navy collar and a pleated navy skirt, worn by <Subject 1>.\n<Subject 3> is a sunny school sports ground with a dirt running track, open blue sky, green trees, and school building in the background.\n<Video 1> is the motion reference video showing a live-action dancing person, providing the dance choreography, rhythm, and body timing.\n\nsummary:\n[reference generation] <Subject 1> performs the full-body dance choreography referenced from <Video 1> while preserving her identity from <Picture 1>, <Picture 2>, and <Picture 3>, wearing <Subject 2> on <Subject 3>.\n\nretention_analysis:\n<Subject 1> (appears in [Shot 1]): fully_preserved - preserve her identity, face, hairstyle, eye shape, and body proportions from the three reference pictures.\n<Subject 2> (appears in [Shot 1]): fully_preserved - keep the summer school uniform throughout the target video.\n<Subject 3> (appears in [Shot 1]): fully_preserved - keep the school ground and outdoor daytime environment throughout the target video.\n<Video 1> (dance motion): transferred - transfer the choreography, pose sequence, body timing, rhythm, and full-body movement into the target anime girl without copying the live-action person's real-life face, clothing, or indoor background.\n\ndetailed_description:\nThe target video is a clean 2D anime-style animation.\n\n[Shot 1] <Subject 1> performs an energetic full-body dance outdoors on <Subject 3>, wearing <Subject 2>. Her sequence of dance poses, arm movements, body rhythm, and timing follow the choreography shown in <Video 1>. The camera maintains a stable full-body framing showing her complete figure from head to toe against the spacious school sports ground under bright daytime sunlight. Her facial appearance, expressive eyes, dark hair, body proportions, summer uniform, and the school ground environment remain visually consistent throughout the clip.\n\noverall_soundscape:\nThere is no dialogue or conversation. Only faint natural outdoor ambient sounds and breeze on the school ground are heard.\n\nnon_diegetic_music:\nThere is no background music (BGM).",
63      "width": 704,
64      "height": 1280,
65      "length": 124,
66      "ref_image_size": "match",
67      "clip": [
68        "10",
69        0
70      ],
71      "vae": [
72        "22",
73        0
74      ],
75      "audio_vae": [
76        "23",
77        0
78      ],
79      "ref_images.ref_image_0": [
80        "11",
81        0
82      ],
83      "ref_images.ref_image_1": [
84        "24",
85        0
86      ],
87      "ref_images.ref_image_2": [
88        "25",
89        0
90      ],
91      "ref_videos.ref_video_0": [
92        "32",
93        0
94      ]
95    },
96    "class_type": "MiniMaxH3ReferenceToVideo",
97    "_meta": {
98      "title": "MiniMax H3 Reference to Video"
99    }
100  },
101  "13": {
102    "inputs": {
103      "noise_seed": 45
104    },
105    "class_type": "RandomNoise",
106    "_meta": {
107      "title": "ランダムノイズ"
108    }
109  },
110  "14": {
111    "inputs": {
112      "sampler_name": "euler"
113    },
114    "class_type": "KSamplerSelect",
115    "_meta": {
116      "title": "Kサンプラー選択"
117    }
118  },
119  "15": {
120    "inputs": {
121      "scheduler": "simple",
122      "steps": 4,
123      "denoise": 1,
124      "model": [
125        "28",
126        0
127      ]
128    },
129    "class_type": "BasicScheduler",
130    "_meta": {
131      "title": "基本スケジューラー"
132    }
133  },
134  "16": {
135    "inputs": {
136      "model": [
137        "28",
138        0
139      ],
140      "conditioning": [
141        "12",
142        0
143      ]
144    },
145    "class_type": "BasicGuider",
146    "_meta": {
147      "title": "基本ガイダー"
148    }
149  },
150  "17": {
151    "inputs": {
152      "noise": [
153        "13",
154        0
155      ],
156      "guider": [
157        "16",
158        0
159      ],
160      "sampler": [
161        "14",
162        0
163      ],
164      "sigmas": [
165        "15",
166        0
167      ],
168      "latent_image": [
169        "12",
170        1
171      ]
172    },
173    "class_type": "SamplerCustomAdvanced",
174    "_meta": {
175      "title": "カスタムサンプラー（高度）"
176    }
177  },
178  "18": {
179    "inputs": {
180      "samples": [
181        "17",
182        0
183      ],
184      "vae": [
185        "22",
186        0
187      ]
188    },
189    "class_type": "VAEDecode",
190    "_meta": {
191      "title": "VAEデコード"
192    }
193  },
194  "19": {
195    "inputs": {
196      "samples": [
197        "17",
198        0
199      ],
200      "vae": [
201        "23",
202        0
203      ]
204    },
205    "class_type": "VAEDecodeAudio",
206    "_meta": {
207      "title": "VAEデコード音声"
208    }
209  },
210  "20": {
211    "inputs": {
212      "fps": 24,
213      "bit_depth": 8,
214      "images": [
215        "18",
216        0
217      ],
218      "audio": [
219        "19",
220        0
221      ]
222    },
223    "class_type": "CreateVideo",
224    "_meta": {
225      "title": "動画を作成"
226    }
227  },
228  "21": {
229    "inputs": {
230      "filename_prefix": "phaseI_diverse_inputs/videoref_from_realperson_seed45",
231      "format": "auto",
232      "codec": "auto",
233      "video": [
234        "20",
235        0
236      ]
237    },
238    "class_type": "SaveVideo",
239    "_meta": {
240      "title": "ビデオを保存"
241    }
242  },
243  "22": {
244    "inputs": {
245      "vae_name": "minimax_h3_video_vae_int8_convrot.safetensors"
246    },
247    "class_type": "VAELoader",
248    "_meta": {
249      "title": "VAEを読み込む"
250    }
251  },
252  "23": {
253    "inputs": {
254      "vae_name": "minimax_h3_audio_vae_fp32.safetensors"
255    },
256    "class_type": "VAELoader",
257    "_meta": {
258      "title": "VAEを読み込む"
259    }
260  },
261  "24": {
262    "inputs": {
263      "image": "testS2_T03_ref_02_upper_528x768.webp"
264    },
265    "class_type": "LoadImage",
266    "_meta": {
267      "title": "画像を読み込む"
268    }
269  },
270  "25": {
271    "inputs": {
272      "image": "testS2_T03_ref_03_face_528x768.webp"
273    },
274    "class_type": "LoadImage",
275    "_meta": {
276      "title": "画像を読み込む"
277    }
278  },
279  "26": {
280    "inputs": {
281      "head_chunks": 4,
282      "model": [
283        "8",
284        0
285      ]
286    },
287    "class_type": "MiniMaxLowVRAMAttention",
288    "_meta": {
289      "title": "MiniMax H3 Low VRAM Attention"
290    }
291  },
292  "28": {
293    "inputs": {
294      "chunks": 2,
295      "seq_threshold": 4096,
296      "model": [
297        "26",
298        0
299      ]
300    },
301    "class_type": "MiniMaxChunkFeedForward",
302    "_meta": {
303      "title": "MiniMax H3 Chunk FeedForward"
304    }
305  },
306  "31": {
307    "inputs": {
308      "file": "INPUT_REAL_PERSON_DANCE_124f.mp4"
309    },
310    "class_type": "LoadVideo",
311    "_meta": {
312      "title": "Load Pose Video"
313    }
314  },
315  "32": {
316    "inputs": {
317      "video": [
318        "31",
319        0
320      ]
321    },
322    "class_type": "GetVideoComponents",
323    "_meta": {
324      "title": "Get Video Components"
325    }
326  }
327}