250 lines
17 KiB
JSON
250 lines
17 KiB
JSON
{
|
|
"1": {
|
|
"inputs": {
|
|
"unet_name": "MiniMax/minimax_h3_ref2va_pruned_int8_convrot.safetensors",
|
|
"weight_dtype": "default"
|
|
},
|
|
"class_type": "UNETLoader",
|
|
"_meta": {
|
|
"title": "UNet加载器"
|
|
}
|
|
},
|
|
"2": {
|
|
"inputs": {
|
|
"clip_name": "MiniMax/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
|
|
"type": "minimax",
|
|
"device": "default"
|
|
},
|
|
"class_type": "CLIPLoader",
|
|
"_meta": {
|
|
"title": "加载CLIP"
|
|
}
|
|
},
|
|
"3": {
|
|
"inputs": {
|
|
"vae_name": "MiniMax/minimax_h3_video_vae_fp16.safetensors"
|
|
},
|
|
"class_type": "VAELoader",
|
|
"_meta": {
|
|
"title": "加载VAE"
|
|
}
|
|
},
|
|
"4": {
|
|
"inputs": {
|
|
"vae_name": "MiniMax/minimax_h3_audio_vae_fp32.safetensors"
|
|
},
|
|
"class_type": "VAELoader",
|
|
"_meta": {
|
|
"title": "加载VAE"
|
|
}
|
|
},
|
|
"5": {
|
|
"inputs": {
|
|
"scheduler": "simple",
|
|
"steps": 8,
|
|
"denoise": 1,
|
|
"model": [
|
|
"22",
|
|
0
|
|
]
|
|
},
|
|
"class_type": "BasicScheduler",
|
|
"_meta": {
|
|
"title": "基本调度器"
|
|
}
|
|
},
|
|
"6": {
|
|
"inputs": {
|
|
"sampler_name": "euler"
|
|
},
|
|
"class_type": "KSamplerSelect",
|
|
"_meta": {
|
|
"title": "K采样器选择"
|
|
}
|
|
},
|
|
"7": {
|
|
"inputs": {
|
|
"width": [
|
|
"24",
|
|
0
|
|
],
|
|
"height": [
|
|
"24",
|
|
1
|
|
],
|
|
"generation_seconds": 11.5,
|
|
"timeline_data": "{\"version\":9,\"fps\":24,\"globalPrompt\":\"\",\"secondPass\":true,\"secondPassModel\":\"minimax_h3_latent_upscaler_3d_conv_v1_fp16.safetensors\",\"secondPassHighSteps\":2,\"selection\":{\"start\":8.5,\"duration\":11.541666666666666},\"videoAudioEnabled\":true,\"videoClips\":[],\"images\":[{\"id\":\"mu5mo29c-dme5zh\",\"file\":\"minimax_h3_timeline_director/1789655485241_lz840_20260324_215319_Z-Image_00001_.png\",\"name\":\"20260324_215319_Z-Image_00001_.png\",\"width\":1080,\"height\":1920},{\"id\":\"mu5moaxk-7a9fy1\",\"file\":\"minimax_h3_timeline_director/1789655496464_mnjqz_20260324_214939_Z-Image_00001_.png\",\"name\":\"20260324_214939_Z-Image_00001_.png\",\"width\":1080,\"height\":1920}],\"audios\":[],\"segmentConfig\":{\"count\":2,\"activeIndex\":1,\"mode\":\"timeline\",\"segments\":[{\"images\":[\"mu5mo29c-dme5zh\"],\"audios\":[],\"prompt\":\"subject_definitions:\\n<Subject 1> is the yellow-furred bear in<Picture 1>, with light golden-yellow fur, a brown military-style canvas jacket with brass buttons and worn shoulder patches, round blue eyes, a large dark-brown nose, and a tense furious expression.\\n<Subject 2> is the red-furred bear in<Picture 2>, with deep russet-red fur, a red traditional Chinese-collar tunic with knotted frog closures down the front, round green eyes, a large dark-red nose, and a cold rage expression.\\n<Audio 1> is the voice-timbre reference for<Subject 1> (S1), containing the spoken Chinese vocal layer referenced by the user.\\n<Audio 2> is the voice-timbre reference for<Subject 2> (S2), containing the spoken Chinese vocal layer referenced by the user.\\n\\nsummary:\\n[reference generation + audio reference] The target video is a 10-second live-action cinematic confrontation clip in which<Subject 1> and<Subject 2> argue fiercely inside a dimly lit room. The five-shot sequence escalates from a wide over-shoulder establishing shot of the two facing each other, to tight close-ups of each bear shouting, to a face-to-face confrontation close-up, ending on a fist slamming onto a wooden table.<Audio 1> and<Audio 2> serve as voice-timbre references for<Subject 1> and<Subject 2>'s Chinese dialogue without copying the original signals.\\n\\nretention_analysis:\\n<Subject 1> (appears in [Shot 1], [Shot 2], [Shot 4], [Shot 5]): fully_preserved - the yellow bear's light golden-yellow fur, brown military-style canvas jacket, round blue eyes, dark-brown nose, and angry expression are retained across all shots.\\n<Subject 2> (appears in [Shot 1], [Shot 3], [Shot 4], [Shot 5]): fully_preserved - the red bear's deep russet-red fur, red traditional Chinese-collar tunic, round green eyes, dark-red nose, and cold rage expression are retained across all shots.\\n<Audio 1>: reference - the target speaker follows<Audio 1>'s voice timbre and emotional intensity without copying the original signal.\\n<Audio 2>: reference - the target speaker follows<Audio 2>'s voice timbre and emotional intensity without copying the original signal.\\n\\ndetailed_description:\\nThe target video uses a 3D CG cinematic style with dramatic chiaroscuro lighting, a dimly lit cluttered old room with peeling walls, scattered books and broken furniture, and a tense confrontation pacing.\\n\\n[Shot 1] A medium over-shoulder shot from behind<Subject 1> frames<Subject 2> standing in the center of a cluttered old room, lit by a single harsh overhead light.<Subject 2>'s deep russet-red fur and red traditional Chinese-collar tunic stand out against the dim background; his round green eyes are narrowed, his large dark-red nose wrinkled in a snarl, his fists clenched at his sides. Across the room, the back of<Subject 1>'s brown military-style canvas jacket is visible in the foreground. The camera holds a static shot with subtle low-amplitude shake as the two bears stare at each other, the air thick with tension.\\n\\n[Shot 2] At 00:02.000, the camera cuts to a tight close-up of<Subject 1>'s face, the cluttered room blurred behind him. The camera pushes in with small amplitude at fast speed toward his face as his round blue eyes lock forward in fury, his dark-brown nose scrunching, his lips pulled back in a snarl, the fur along his neck bristling. He speaks in a deep, gravelly adult male voice with a fierce aggressive rising cadence, citing voice timbre from<Audio 1> (S1), <d>[Chinese] 你以为这样就能赢我吗!</d> His breath comes fast, his nostrils flaring as the words land.\\n\\n[Shot 3] At 00:04.000, the camera cuts to a tight close-up of<Subject 2>'s face, the dim room softly blurred behind him. The camera pushes in with small amplitude at fast speed toward his face as his round green eyes narrow into slits, his large dark-red nose curls in contempt, his teeth bared. He speaks in a deeper, hoarse adult male voice with a cold seething cadence, citing voice timbre from<Audio 2> (S2), <d>[Chinese] 我受够了!</d> His throat vibrates with the low roar, the fur along his cheeks rippling with the force of the shout.\\n\\n[Shot 4] At 00:06.000, the camera cuts to a tight two-shot, framing both bears forehead-to-forehead at close range, their faces filling the left and right halves of the frame. The camera tracks left and right with medium amplitude at fast speed as both bears roar together, heads jerking forward, the camera shaking slightly with the force of their voices. The yellow fur of<Subject 1> and the red fur of<Subject 2> blur at the edges of the frame, a single droplet of sweat trembling on the tip of<Subject 1>'s nose.\\n\\n[Shot 5] At 00:08.000, the camera cuts to an extreme close-up of<Subject 1>'s clenched paw slamming down onto the wooden table, the impact sending a chipped ceramic teacup flying, dark tea splashing upward, a stack of paper scattering. The camera shakes strongly with small amplitude at slow speed as the paw lands, the wood cracking under the blow, the final frame freezing on the shattered teacup and the yellow paw planted on the wood.\\n\\noverall_soundscape:\\nTwo bears roar and shout aggressively at each other inside the enclosed room, their voices overlapping in a rising swell of fury. A wooden floor creaks under shifting weight, fabric tears at one seam, a ceramic teacup cracks and shatters against the table, dark liquid splashes, and the impact reverberates through the wood. A low metallic echo lingers in the room as the final frame freezes.\\n\\nnon_diegetic_music:\\nA driving low brass pulse at a fast tempo with sharp staccato cello stabs on every shout, joined by a heavy percussion hit on the table slam that decays into a sustained low cello drone, fading out as the final frame freezes.\",\"startFrame\":0,\"endFrame\":243},{\"images\":[\"mu5moaxk-7a9fy1\"],\"audios\":[],\"prompt\":\"subject_definitions:\\n<Subject 1> is the yellow-furred bear in<Picture 1>, with light golden-yellow fur, a brown military-style canvas jacket with brass buttons and worn shoulder patches, round blue eyes, a large dark-brown nose, and a tense furious expression.\\n<Subject 2> is the red-furred bear in<Picture 2>, with deep russet-red fur, a red traditional Chinese-collar tunic with knotted frog closures down the front, round green eyes, a large dark-red nose, and a cold rage expression.\\n<Audio 1> is the voice-timbre reference for<Subject 1> (S1), containing the spoken Chinese vocal layer referenced by the user.\\n<Audio 2> is the voice-timbre reference for<Subject 2> (S2), containing the spoken Chinese vocal layer referenced by the user.\\n\\nsummary:\\n[reference generation + audio reference] The target video is a 10-second live-action cinematic confrontation clip in which<Subject 1> and<Subject 2> argue fiercely inside a dimly lit room. The five-shot sequence escalates from a wide over-shoulder establishing shot of the two facing each other, to tight close-ups of each bear shouting, to a face-to-face confrontation close-up, ending on a fist slamming onto a wooden table.<Audio 1> and<Audio 2> serve as voice-timbre references for<Subject 1> and<Subject 2>'s Chinese dialogue without copying the original signals.\\n\\nretention_analysis:\\n<Subject 1> (appears in [Shot 1], [Shot 2], [Shot 4], [Shot 5]): fully_preserved - the yellow bear's light golden-yellow fur, brown military-style canvas jacket, round blue eyes, dark-brown nose, and angry expression are retained across all shots.\\n<Subject 2> (appears in [Shot 1], [Shot 3], [Shot 4], [Shot 5]): fully_preserved - the red bear's deep russet-red fur, red traditional Chinese-collar tunic, round green eyes, dark-red nose, and cold rage expression are retained across all shots.\\n<Audio 1>: reference - the target speaker follows<Audio 1>'s voice timbre and emotional intensity without copying the original signal.\\n<Audio 2>: reference - the target speaker follows<Audio 2>'s voice timbre and emotional intensity without copying the original signal.\\n\\ndetailed_description:\\nThe target video uses a 3D CG cinematic style with dramatic chiaroscuro lighting, a dimly lit cluttered old room with peeling walls, scattered books and broken furniture, and a tense confrontation pacing.\\n\\n[Shot 1] A medium over-shoulder shot from behind<Subject 1> frames<Subject 2> standing in the center of a cluttered old room, lit by a single harsh overhead light.<Subject 2>'s deep russet-red fur and red traditional Chinese-collar tunic stand out against the dim background; his round green eyes are narrowed, his large dark-red nose wrinkled in a snarl, his fists clenched at his sides. Across the room, the back of<Subject 1>'s brown military-style canvas jacket is visible in the foreground. The camera holds a static shot with subtle low-amplitude shake as the two bears stare at each other, the air thick with tension.\\n\\n[Shot 2] At 00:02.000, the camera cuts to a tight close-up of<Subject 1>'s face, the cluttered room blurred behind him. The camera pushes in with small amplitude at fast speed toward his face as his round blue eyes lock forward in fury, his dark-brown nose scrunching, his lips pulled back in a snarl, the fur along his neck bristling. He speaks in a deep, gravelly adult male voice with a fierce aggressive rising cadence, citing voice timbre from<Audio 1> (S1), <d>[Chinese] 你以为这样就能赢我吗!</d> His breath comes fast, his nostrils flaring as the words land.\\n\\n[Shot 3] At 00:04.000, the camera cuts to a tight close-up of<Subject 2>'s face, the dim room softly blurred behind him. The camera pushes in with small amplitude at fast speed toward his face as his round green eyes narrow into slits, his large dark-red nose curls in contempt, his teeth bared. He speaks in a deeper, hoarse adult male voice with a cold seething cadence, citing voice timbre from<Audio 2> (S2), <d>[Chinese] 我受够了!</d> His throat vibrates with the low roar, the fur along his cheeks rippling with the force of the shout.\\n\\n[Shot 4] At 00:06.000, the camera cuts to a tight two-shot, framing both bears forehead-to-forehead at close range, their faces filling the left and right halves of the frame. The camera tracks left and right with medium amplitude at fast speed as both bears roar together, heads jerking forward, the camera shaking slightly with the force of their voices. The yellow fur of<Subject 1> and the red fur of<Subject 2> blur at the edges of the frame, a single droplet of sweat trembling on the tip of<Subject 1>'s nose.\\n\\n[Shot 5] At 00:08.000, the camera cuts to an extreme close-up of<Subject 1>'s clenched paw slamming down onto the wooden table, the impact sending a chipped ceramic teacup flying, dark tea splashing upward, a stack of paper scattering. The camera shakes strongly with small amplitude at slow speed as the paw lands, the wood cracking under the blow, the final frame freezing on the shattered teacup and the yellow paw planted on the wood.\\n\\noverall_soundscape:\\nTwo bears roar and shout aggressively at each other inside the enclosed room, their voices overlapping in a rising swell of fury. A wooden floor creaks under shifting weight, fabric tears at one seam, a ceramic teacup cracks and shatters against the table, dark liquid splashes, and the impact reverberates through the wood. A low metallic echo lingers in the room as the final frame freezes.\\n\\nnon_diegetic_music:\\nA driving low brass pulse at a fast tempo with sharp staccato cello stabs on every shout, joined by a heavy percussion hit on the table slam that decays into a sustained low cello drone, fading out as the final frame freezes.\",\"startFrame\":204,\"endFrame\":481}]}}"
|
|
},
|
|
"class_type": "MiniMaxH3TimelinePlanner",
|
|
"_meta": {
|
|
"title": "MiniMax H3 素材规划台"
|
|
}
|
|
},
|
|
"8": {
|
|
"inputs": {
|
|
"seed": 42,
|
|
"continue_audio_latent": true,
|
|
"ref_image_size": "match",
|
|
"model": [
|
|
"22",
|
|
0
|
|
],
|
|
"clip": [
|
|
"2",
|
|
0
|
|
],
|
|
"vae": [
|
|
"3",
|
|
0
|
|
],
|
|
"audio_vae": [
|
|
"4",
|
|
0
|
|
],
|
|
"finite_plan": [
|
|
"7",
|
|
2
|
|
],
|
|
"sampler": [
|
|
"6",
|
|
0
|
|
],
|
|
"sigmas": [
|
|
"5",
|
|
0
|
|
]
|
|
},
|
|
"class_type": "MiniMaxH3FiniteSegmentSampler",
|
|
"_meta": {
|
|
"title": "MiniMax H3 有限分段采样"
|
|
}
|
|
},
|
|
"9": {
|
|
"inputs": {
|
|
"fps": 24,
|
|
"bit_depth": 8,
|
|
"color_space": "sRGB",
|
|
"codec": "none",
|
|
"images": [
|
|
"8",
|
|
1
|
|
],
|
|
"audio": [
|
|
"8",
|
|
2
|
|
]
|
|
},
|
|
"class_type": "CreateVideo",
|
|
"_meta": {
|
|
"title": "创建视频"
|
|
}
|
|
},
|
|
"10": {
|
|
"inputs": {
|
|
"filename_prefix": "H3_plugin_matrix/CASE_ID",
|
|
"format": "mp4",
|
|
"format.codec": "auto",
|
|
"codec": "h264",
|
|
"codec.encoding": "auto",
|
|
"video": [
|
|
"9",
|
|
0
|
|
]
|
|
},
|
|
"class_type": "SaveVideo",
|
|
"_meta": {
|
|
"title": "保存视频"
|
|
}
|
|
},
|
|
"18": {
|
|
"inputs": {
|
|
"shift_video": 10,
|
|
"shift_audio": 3,
|
|
"model": [
|
|
"21",
|
|
0
|
|
]
|
|
},
|
|
"class_type": "MiniMaxH3SigmaShift",
|
|
"_meta": {
|
|
"title": "MiniMax H3 Sigma Shift"
|
|
}
|
|
},
|
|
"19": {
|
|
"inputs": {
|
|
"attention": "comfy kitchen attention",
|
|
"model": [
|
|
"18",
|
|
0
|
|
]
|
|
},
|
|
"class_type": "ModelAttentionBackend",
|
|
"_meta": {
|
|
"title": "模型注意力后端"
|
|
}
|
|
},
|
|
"20": {
|
|
"inputs": {
|
|
"selection": "vsa",
|
|
"selection.keep_percent": 10,
|
|
"start_percent": 0.2,
|
|
"end_percent": 1,
|
|
"dense_blocks": "",
|
|
"min_tokens": 12288,
|
|
"extra_tokens": 256,
|
|
"sink_conditioning": "exact_kv_and_rows",
|
|
"verbose": false,
|
|
"model": [
|
|
"19",
|
|
0
|
|
]
|
|
},
|
|
"class_type": "BlockSparseAttention",
|
|
"_meta": {
|
|
"title": "模型稀疏注意力"
|
|
}
|
|
},
|
|
"21": {
|
|
"inputs": {
|
|
"lora_name": "MiniMax/minimax_h3_ref2v_turbo_8step_v1.0_768p_comfyui_bf16.safetensors",
|
|
"strength_model": 1,
|
|
"model": [
|
|
"1",
|
|
0
|
|
]
|
|
},
|
|
"class_type": "LoraLoaderModelOnly",
|
|
"_meta": {
|
|
"title": "LoRA加载器(仅模型)"
|
|
}
|
|
},
|
|
"22": {
|
|
"inputs": {
|
|
"max_resolution": 1024,
|
|
"jpeg_quality": 80,
|
|
"suppress_default_preview": true,
|
|
"preview_frames": 1,
|
|
"preview_fps": 12,
|
|
"tiny_vae": "taeh3.safetensors",
|
|
"model": [
|
|
"20",
|
|
0
|
|
]
|
|
},
|
|
"class_type": "ModelPreviewOverrideKJ",
|
|
"_meta": {
|
|
"title": "Model Preview Override"
|
|
}
|
|
},
|
|
"24": {
|
|
"inputs": {
|
|
"aspect_ratio": "9:16 (Portrait Widescreen)",
|
|
"megapixels": 0.9,
|
|
"multiple": 32
|
|
},
|
|
"class_type": "ResolutionSelector",
|
|
"_meta": {
|
|
"title": "分辨率选择器"
|
|
}
|
|
}
|
|
} |