Wan Animate 2 : Transfert de Mouvement
Animez un personnage fixe à partir d’une vidéo de pilotage avec Wan Animate 2, en transférant le mouvement directement depuis les frames, sans extraction de pose ni prétraitement de squelette. La sortie préserve l’identité du personnage de référence tout en générant un nouvel arrière-plan et un mouvement de caméra depuis votre prompt texte. Idéal pour le transfert de mouvement, l’imitation d’actions et des projets de remplacement d’animation de personnage.
Dans le workflow
Un aperçu en lecture seule du modèle source. Ouvrez-le dans Cloud pour explorer les nœuds et apporter des modifications.
Loading the graph
Open full-size workflow previewAppelez ce modèle depuis votre code
La requête ci-dessous utilise le même graphe préparé et les mêmes paramètres que le Playground. Chaque envoi peut lancer une exécution payante ; ne relancez pas automatiquement un envoi dont l’issue est inconnue.
What you need
- Runs on
- Comfy Cloud
- Clé
- COMFY_API_KEY
- Vos fichiers
- Importés par le code depuis leurs URL
Nécessite un forfait Cloud payant et des crédits.
# Python 3.10+: pip install comfy-sdk==0.4.0
import os
from comfy_sdk import Comfy
api_key = os.environ.get("COMFY_API_KEY", "").strip()
if not api_key: raise RuntimeError("Set COMFY_API_KEY")
os.environ.setdefault("COMFY_BASE_URL", "https://cloud.comfy.org")
client = Comfy(api_key=api_key)
workflow = client.workflows.from_json({
"189": {
"class_type": "LoadImage",
"inputs": {
"image": "UPLOADED_image_FILENAME"
},
"_meta": {
"title": "Load Image (Reference Image)"
}
},
"240": {
"class_type": "LoadVideo",
"inputs": {
"file": "UPLOADED_video_FILENAME"
},
"_meta": {
"title": "Load Video (Pose Video)"
}
},
"245": {
"class_type": "CreateVideo",
"inputs": {
"audio": [
"288",
1
],
"bit_depth": 8,
"codec": "none",
"color_space": "sRGB",
"fps": [
"288",
2
],
"images": [
"289",
0
]
},
"_meta": {
"title": "Create Video"
}
},
"246": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/ComfyUI",
"format": "auto",
"format.codec": "auto",
"video": [
"245",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"288": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"240",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"289": {
"class_type": "BatchImagesNode",
"inputs": {
"images.image0": [
"261:299",
0
]
},
"_meta": {
"title": "Batch Images"
}
},
"292": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/ComfyUI",
"format": "auto",
"format.codec": "auto",
"video": [
"291:80",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"536": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a / 81",
"values.a": [
"537",
2
]
},
"_meta": {
"title": "Math Expression"
}
},
"537": {
"class_type": "GetImageSize",
"inputs": {
"image": [
"288",
0
]
},
"_meta": {
"title": "Get Image Size"
}
},
"538": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"536",
0
]
},
"_meta": {
"title": "Preview as Text"
}
},
"261:11": {
"class_type": "LoraLoaderModelOnly",
"inputs": {
"lora_name": "lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors",
"model": [
"261:239",
0
],
"strength_model": 1
},
"_meta": {
"title": "Load LoRA"
}
},
"261:18": {
"class_type": "BasicScheduler",
"inputs": {
"denoise": 1,
"model": [
"261:224",
0
],
"scheduler": "simple",
"steps": 6
},
"_meta": {
"title": "BasicScheduler"
}
},
"261:19": {
"class_type": "SamplerCustom",
"inputs": {
"add_noise": True,
"cfg": 1,
"latent_image": [
"261:247",
2
],
"model": [
"261:95",
0
],
"negative": [
"261:247",
1
],
"noise_seed": 951677846216619,
"positive": [
"261:247",
0
],
"sampler": [
"261:27",
0
],
"sigmas": [
"261:18",
0
]
},
"_meta": {
"title": "SamplerCustom"
}
},
"261:220": {
"class_type": "CLIPVisionEncode",
"inputs": {
"clip_vision": [
"261:75",
0
],
"crop": "none",
"image": [
"261:236",
0
]
},
"_meta": {
"title": "CLIP Vision Encode"
}
},
"261:222": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"261:9",
0
],
"text": "An elderly mystic doing an energetic contemporary street dance with rhythmic steps, wide arm swings, a spin and a crouch, background stationary"
},
"_meta": {
"title": "CLIP Text Encode (Prompt)"
}
},
"261:223": {
"class_type": "TrimVideoLatent",
"inputs": {
"samples": [
"261:19",
0
],
"trim_amount": [
"261:247",
3
]
},
"_meta": {
"title": "Trim Video Latent"
}
},
"261:224": {
"class_type": "WanAnimate2Cache",
"inputs": {
"device": "gpu",
"dtype": "int8",
"model": [
"261:258",
0
]
},
"_meta": {
"title": "WanAnimate2Cache"
}
},
"261:236": {
"class_type": "ImageFromBatch",
"inputs": {
"batch_index": 0,
"image": [
"261:243",
0
],
"length": 1
},
"_meta": {
"title": "Get Image from Batch"
}
},
"261:239": {
"class_type": "UNETLoader",
"inputs": {
"unet_name": "wan_animate_2_int8_convrot.safetensors",
"weight_dtype": "default"
},
"_meta": {
"title": "Load Diffusion Model"
}
},
"261:241": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"240",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"261:243": {
"class_type": "ResizeImageMaskNode",
"inputs": {
"input": [
"261:241",
0
],
"resize_type": "scale dimensions",
"resize_type.crop": "center",
"resize_type.height": 480,
"resize_type.width": 832,
"scale_method": "area"
},
"_meta": {
"title": "Resize Image/Mask"
}
},
"261:244": {
"class_type": "ResizeImageMaskNode",
"inputs": {
"input": [
"189",
0
],
"resize_type": "scale dimensions",
"resize_type.crop": "center",
"resize_type.height": [
"261:256",
1
],
"resize_type.width": [
"261:256",
0
],
"scale_method": "area"
},
"_meta": {
"title": "Resize Image/Mask"
}
},
"261:247": {
"class_type": "WanAnimate2ToVideo",
"inputs": {
"batch_size": 1,
"clip_vision_output": [
"261:76",
0
],
"clip_vision_output_pose": [
"261:220",
0
],
"height": [
"261:256",
1
],
"length": 81,
"negative": [
"261:4",
0
],
"pose_end_percent": 1,
"pose_start_percent": 0,
"pose_strength": 1,
"pose_video": [
"261:243",
0
],
"positive": [
"261:3",
0
],
"positive_pose": [
"261:222",
0
],
"reference_image": [
"261:244",
0
],
"reference_image_strength": 1,
"vae": [
"261:7",
0
],
"video_frame_offset": 0,
"width": [
"261:256",
0
]
},
"_meta": {
"title": "WanAnimate2ToVideo"
}
},
"261:256": {
"class_type": "GetImageSize",
"inputs": {
"image": [
"261:243",
0
]
},
"_meta": {
"title": "Get Image Size"
}
},
"261:257": {
"class_type": "ContextWindowsManual",
"inputs": {
"causal_window_fix": True,
"closed_loop": False,
"cond_retain_index_list": "0",
"context_length": 21,
"context_overlap": 8,
"context_schedule": "standard_static",
"context_stride": 1,
"dim": 2,
"freenoise": True,
"fuse_method": "pyramid",
"latent_retain_index_list": "",
"model": [
"261:11",
0
],
"split_conds_to_windows": False
},
"_meta": {
"title": "Context Windows (Manual)"
}
},
"261:258": {
"class_type": "ComfySwitchNode",
"inputs": {
"on_false": [
"261:11",
0
],
"on_true": [
"261:257",
0
],
"switch": False
},
"_meta": {
"title": "If/Else Switch"
}
},
"261:27": {
"class_type": "KSamplerSelect",
"inputs": {
"sampler_name": "lcm"
},
"_meta": {
"title": "KSamplerSelect"
}
},
"261:297": {
"class_type": "PrimitiveBoolean",
"inputs": {
"value": False
},
"_meta": {
"title": "Boolean"
}
},
"261:298": {
"class_type": "ImageFromBatch",
"inputs": {
"batch_index": 1,
"image": [
"261:6",
0
],
"length": 4096
},
"_meta": {
"title": "Get Image from Batch"
}
},
"261:299": {
"class_type": "ComfySwitchNode",
"inputs": {
"on_false": [
"261:6",
0
],
"on_true": [
"261:298",
0
],
"switch": [
"261:297",
0
]
},
"_meta": {
"title": "If/Else Switch"
}
},
"261:3": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"261:9",
0
],
"text": "Character appearance description: A tall, sinewy elderly mystic with deeply creased pale slate-grey skin, intense amber eyes and a long braided steel-grey beard falling over his collarbones. A crown of polished white stag antlers bound with sinew sits on his head. A glossy black raven-feather mantle covers his shoulders over a bare, scarred torso. He wears a heavy pleated crimson wool kilt with a herringbone weave, a belt of carved ivory plaques and tarnished silver rings, and dark sealskin bindings wrapped around his lower legs; his feet are bare.\nBackground description: A plain, dark charcoal backdrop with soft drifting mist and cold blue directional light, no decorations."
},
"_meta": {
"title": "CLIP Text Encode (Positive Prompt)"
}
},
"261:4": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"261:9",
0
],
"text": "色调艳丽,过曝,静态,细节模糊不清,字幕,风格,作品,画作,画面,静止,整体发灰,最差质量,低质量,JPEG压缩残留,丑陋的,残缺的,多余的手指,画得不好的手部,画得不好的脸部,畸形的,毁容的,形态畸形的肢体,手指融合,静止不动的画面,杂乱的背景,三条腿,背景人很多,倒着走"
},
"_meta": {
"title": "CLIP Text Encode (Negative Prompt)"
}
},
"261:6": {
"class_type": "VAEDecode",
"inputs": {
"samples": [
"261:223",
0
],
"vae": [
"261:7",
0
]
},
"_meta": {
"title": "VAE Decode"
}
},
"261:7": {
"class_type": "VAELoader",
"inputs": {
"vae_name": "Wan2_1_VAE_bf16.safetensors"
},
"_meta": {
"title": "Load VAE"
}
},
"261:75": {
"class_type": "CLIPVisionLoader",
"inputs": {
"clip_name": "clip_vision_h.safetensors"
},
"_meta": {
"title": "Load CLIP Vision"
}
},
"261:76": {
"class_type": "CLIPVisionEncode",
"inputs": {
"clip_vision": [
"261:75",
0
],
"crop": "none",
"image": [
"261:244",
0
]
},
"_meta": {
"title": "CLIP Vision Encode"
}
},
"261:9": {
"class_type": "CLIPLoader",
"inputs": {
"clip_name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
"device": "default",
"type": "wan"
},
"_meta": {
"title": "Load CLIP"
}
},
"261:95": {
"class_type": "ModelSamplingSD3",
"inputs": {
"model": [
"261:224",
0
],
"shift": 5
},
"_meta": {
"title": "ModelSamplingSD3"
}
},
"291:290": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a & ~1",
"values.a": [
"291:90",
0
]
},
"_meta": {
"title": "Math Expression"
}
},
"291:416": {
"class_type": "GetImageSize",
"inputs": {
"image": [
"291:77",
0
]
},
"_meta": {
"title": "Get Image Size"
}
},
"291:417": {
"class_type": "ImageFromBatch",
"inputs": {
"batch_index": 0,
"image": [
"291:77",
0
],
"length": [
"291:572",
1
]
},
"_meta": {
"title": "Get Image from Batch"
}
},
"291:571": {
"class_type": "GetImageSize",
"inputs": {
"image": [
"291:78",
0
]
},
"_meta": {
"title": "Get Image Size"
}
},
"291:572": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "min(a,b)",
"values.a": [
"291:416",
2
],
"values.b": [
"291:571",
2
]
},
"_meta": {
"title": "Math Expression"
}
},
"291:573": {
"class_type": "ImageFromBatch",
"inputs": {
"batch_index": 0,
"image": [
"291:78",
0
],
"length": [
"291:572",
1
]
},
"_meta": {
"title": "Get Image from Batch"
}
},
"291:77": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"246",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"291:78": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"240",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"291:79": {
"class_type": "ImageStitch",
"inputs": {
"direction": "right",
"image1": [
"291:417",
0
],
"image2": [
"291:573",
0
],
"match_image_size": True,
"spacing_color": "white",
"spacing_width": 0
},
"_meta": {
"title": "Stitch Images"
}
},
"291:80": {
"class_type": "CreateVideo",
"inputs": {
"audio": [
"291:77",
1
],
"bit_depth": 8,
"codec": "none",
"color_space": "sRGB",
"fps": [
"291:77",
2
],
"images": [
"291:97",
0
]
},
"_meta": {
"title": "Create Video"
}
},
"291:90": {
"class_type": "GetImageSize",
"inputs": {
"image": [
"291:79",
0
]
},
"_meta": {
"title": "Get Image Size"
}
},
"291:96": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a & ~1",
"values.a": [
"291:90",
1
]
},
"_meta": {
"title": "Math Expression"
}
},
"291:97": {
"class_type": "ResizeImageMaskNode",
"inputs": {
"input": [
"291:79",
0
],
"resize_type": "scale dimensions",
"resize_type.crop": "center",
"resize_type.height": [
"291:96",
1
],
"resize_type.width": [
"291:290",
1
],
"scale_method": "area"
},
"_meta": {
"title": "Resize Image/Mask"
}
}
})
workflow.set_input("189", "image", client.assets.from_url("https://comfy.org/catalog-media/daad57c5-e3b2-44b3-a5e6-7b9ccd2801af/8aa22c70-41c6-4a3a-ae58-d675bb83f3fe"))
workflow.set_input("240", "file", client.assets.from_url("https://comfy.org/catalog-media/7efe8b53-ab50-4c05-b6f5-17cf11809cd4/09024a32-55f2-4c41-8582-e1497516ae1d"))
# run() also passes api_key to the partner nodes in this workflow.
job = client.run(workflow, api_key=api_key)
for output in job.get_outputs("246"):
output.to_file(output.name)
for output in job.get_outputs("292"):
output.to_file(output.name)Essayer un exemple
À propos de ce workflow
Ce workflow ComfyUI montre un transfert de mouvement de bout en bout avec Wan-Animate-2 (par Wan-Video/Alibaba) : une seule image de référence (votre personnage) est animée pour suivre le mouvement d’une vidéo de pilotage, sans extraction de points clés de pose ni de squelette. Le cœur du graphe est le sous-graphe « Motion Transfer (Wan Animate 2) » alimenté par des nœuds personnalisés Wan-Animate-2 (UUID : 11706f8a-428d-4ef9-b24f-863f651c1b0b, 79280513-53b2-4417-93f8-f3e116bb6736, b798a87e-23ec-483d-ab9b-6af5b298d0ad). Ces nœuds prennent votre image de référence, le signal de mouvement par frame de la vidéo de pilotage, et un prompt texte qui guide l’arrière-plan et la sensation de caméra. Comme le mouvement est appris directement à partir des frames, la mise en place est rapide et robuste — sans OpenPose ni passes de prétraitement.
LoadImage fournit l’identité du personnage, tandis que LoadVideo et GetVideoComponents extraient les frames et le fps. BatchImagesNode duplique l’image de référence pour que chaque frame vidéo ait sa propre entrée d’identité associée. GetImageSize et ComfyMathExpression servent à garder une résolution et un format cohérents entre entrées et sorties. PreviewAny permet d’inspecter des frames intermédiaires, et CreateVideo + SaveVideo réassemblent les frames générées en vidéo en utilisant le fps d’origine. Pour les clips longs, vous dupliquez manuellement le sous-graphe Motion Transfer (la prise en charge du bouclage natif est en attente upstream) et vous les connectez en série pour que la chaîne puisse traiter toute la vidéo de pilotage.
Questions-réponses
Non. Wan-Animate-2 effectue un transfert de mouvement direct depuis les images de la vidéo de pilotage. Le flux de travail ne nécessite pas de keypoints ni de prétraitement de pose.
Utilisez une image de référence propre, bien éclairée, avec peu d’occlusion et un arrière-plan simple. Gardez le ratio d’aspect cohérent et évitez les prompts qui décrivent un personnage ou des traits du visage différents et en conflit avec votre référence.
GetVideoComponents fournit les fps du clip de pilotage — transmettez-les à CreateVideo pour que la lecture corresponde au timing source. Pour retimer, ajustez les fps dans CreateVideo (fps plus élevé = lecture plus rapide, fps plus bas = lecture plus lente).
Le support natif de boucle est encore en cours d’évaluation. Pour l’instant, dupliquez le sous-graphe « Motion Transfer (Wan Animate 2) » : activez la copie contournée (Ctrl+B), assurez-vous que son prompt et sa taille correspondent au premier, connectez sa Sortie d’image à BatchImagesNode, puis copiez-collez (Ctrl+C, Ctrl+Shift+V) assez de segments pour couvrir toute la vidéo de pilotage.