WAN LongCatAvatar: Create Talking AI Avatars in ComfyUI

About This Workflow

Create talking avatar videos from a single image using audio or text input.

Generate natural head movement, lip-sync, and facial expressions while keeping the face identity consistent.

Animate human, anime, or stylized characters for shorts, explainers, or AI presenters.

Combine LongCatAvatar with voice models, TTS, or audio-driven workflows inside ComfyUI.

Run the full avatar generation locally, giving better control, privacy, and customization.

Required Models

diffusion model

• fp8 KJ: LongCat-Avatar-single_fp8_e4m3fn_scaled_mixed_KJ.safetensors
• bf16 full model: [LongCat-Avatar_comfy_bf16.safetensors] https://huggingface.co/Kijai/LongCat-Video_comfy/blob/main/Avatar/LongCat-Avatar_comfy_bf16.safetensors

** LORA **

• LoRA KJ: LongCat_distill_lora_alpha64_bf16.safetensors

vae

• wan_2.1_vae.safetensors

clip vision

• clip_vision_h.safetensors

text encoder

• umt5-xxl-enc-bf16.safetensors

Model Storage Location

📂 ComfyUI/
├── 📂 models/
│   ├── 📂 diffusion_models/
│   │   └── LongCat-Avatar-single_fp8_e4m3fn_scaled_mixed_KJ.safetensors or LongCat-Avatar_comfy_bf16.safetensors
│   ├── 📂 vae/
│   │   └── wan_2.1_vae.safetensors
│   ├── 📂 lora/
│   │   └── LongCat_distill_lora_alpha64_bf16.safetensors
│   ├── 📂 clip_vision/
│   │   └── clip_vision_h.safetensors
│   └── 📂 text_encoders/
│       └── umt5-xxl-enc-bf16.safetensors

Workflow Graph