---
license: apache-2.0
tags: [comfyui, infinitetalk, wan2.1, talking-head]
---

# InfiniteTalk models for Talk Studio

The files ComfyUI's native InfiniteTalk workflow needs, gathered into one repo so a RunPod serverless endpoint
can cache them. Copied unchanged from the original repos:

| File | From | License |
|---|---|---|
| `diffusion_models/Wan2_1-I2V-14B-480p_fp8_e4m3fn_scaled_KJ.safetensors` | [Kijai/WanVideo_comfy_fp8_scaled](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled) | Apache 2.0 |
| `loras/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors` | [Kijai/WanVideo_comfy](https://huggingface.co/Kijai/WanVideo_comfy) | Apache 2.0 |
| `text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors` | [Comfy-Org/Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) | Apache 2.0 |
| `vae/Wan2_1_VAE_bf16.safetensors` | [Kijai/WanVideo_comfy](https://huggingface.co/Kijai/WanVideo_comfy) | Apache 2.0 |
| `model_patches/wan2.1_infiniteTalk_single_fp16.safetensors` | [Comfy-Org/Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) | Apache 2.0 |
| `audio_encoders/wav2vec2-chinese-base_fp16.safetensors` | [Kijai/wav2vec2_safetensors](https://huggingface.co/Kijai/wav2vec2_safetensors) | MIT |

Wan2.1, InfiniteTalk and the lightx2v LoRA are Apache 2.0; the wav2vec2 audio encoder (TencentGameMate/chinese-wav2vec2-base) is MIT.