Wan 2.1 for ComfyUI

Wan 2.1 for ComfyUI

Repackaged Wan 2.1 video model files for ComfyUI, covering text to video, image to video, VACE editing, camera control, alpha video and audio-driven dubbing, with ready workflows.

Made by Comfy-Org on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: These are Wan 2.1 video model files repackaged for ComfyUI, a node-based tool. One repository gathers many variants: text to video, image to video, first and last frame, Fun control and camera models, VACE, MAGREF, Wan-Alpha and InfiniteTalk. It also holds the text encoders, VAEs and a CLIP vision file they need, and the card lists 15 ComfyUI workflows.

Who it suits: It suits ComfyUI users who make AI video and want one place to get the Wan 2.1 files in the layout ComfyUI expects.

Industry | Who uses it | How they use it | Example
Film and video | Video artists | Turn a text prompt or a still image into a clip | The Text to Video and Image to Video workflows
Marketing | Content teams | Make short clips and add speech to a video | The InfiniteTalk workflow for audio-driven full-body dubbing
Post-production | Editors | Fill, extend or guide video with VACE | The VACE inpainting, outpainting and reference-to-video workflows

You can skip it if you do not use ComfyUI, since these files are packaged for it.

Where to use it:

• ComfyUI, on Windows, macOS or Linux
• The ComfyUI model folders: clip_vision, diffusion_models, loras, model_patches, text_encoders and vae
• Comfy-Org workflow templates, which the card lists by name
• Python with the Hugging Face tools, to download the files

Use case | Who | What they get back
Text or image to video | Video artists | A generated clip from a prompt or a starting image
Video editing with VACE | Editors | Inpainted, outpainted or reference-guided video
Speech-driven video | Content teams | A video dubbed from audio through InfiniteTalk

How to set it up: You need ComfyUI and room for large video model files.

• A working ComfyUI install
• Python with the huggingface_hub package
• Plenty of disk space, since the repository holds many large files

1. Install the Hugging Face tools with Python.
2. Download the repository files with the second command below.
3. Place the files you need in the matching folders under ComfyUI/models.
4. Open a listed workflow in ComfyUI and select the files it needs.

pip install -U huggingface_hub
hf download Comfy-Org/Wan-2.1

Title: Where the files go under ComfyUI/models
Folder | What goes there
clip_vision | clip_vision_h.safetensors
diffusion_models | The Wan 2.1 video models, in fp16, bf16 and fp8 forms
loras | wan_alpha_2.1_rgba_lora.safetensors
model_patches | Two InfiniteTalk files, one for a single speaker and one for several
text_encoders | umt5_xxl_fp16.safetensors and umt5_xxl_fp8_e4m3fn_scaled.safetensors
vae | wan_2.1_vae.safetensors and two Wan-Alpha files for the alpha and rgb channels

How to use it: The card gives no command. Open one of the listed workflows in ComfyUI, load the matching model files, type your prompt or load your input, and run it.

Title: The model families in the diffusion_models folder
Family | Sizes | What it does
Text to video | 1.3B and 14B | Makes a clip from a text prompt
Image to video | 14B at 480P and 720P | Animates a still image
First and last frame | 14B at 720P | Builds a clip between two given frames
Fun control and camera | 1.3B and 14B | Guides a clip with control input or camera moves
Fun inpainting | 1.3B | Fills in parts of a video
VACE | 1.3B and 14B | Editing tasks such as inpainting, outpainting and reference to video
MAGREF | 14B | Listed in the files, with no workflow named on the card

Title: Some of the listed workflows
Tool | What it does
Wan 2.1 Text to Video | Generates a clip from a text prompt
Wan 2.1 Image to Video | Animates a still image
Wan 2.1 Inpainting | Fills in parts of a video
Wan 2.1 ControlNet | Guides a clip with control input
Wan 2.1 Fun Camera 1.3B | Camera-guided video with the smaller model
Wan2.1 VACE Control Video | Video-to-video editing with VACE
Wan2.1 VACE Outpainting | Extends the frame of a video
Wan2.1 VACE Reference to Video | Uses a reference to guide the clip
Wan2.1 Alpha T2V | Text to video with an alpha channel
InfiniteTalk | Audio-driven full-body video dubbing

Example: You place the text encoder, the VAE and one text-to-video model file in their folders. Then you open the Wan 2.1 Text to Video workflow, type a short description of a scene, and run it. ComfyUI returns a short generated clip.

For newcomers and experts: Think of it as a toolbox of video generators for ComfyUI. A first small try is the Text to Video workflow with a 1.3B model file. The card does not say which files each workflow expects, so check the workflow's own notes.

Under the hood, each task has its own diffusion model, offered in several precisions such as fp16, bf16, fp8 e4m3fn and fp8 scaled. Two text encoders and a shared VAE serve most of them. Wan-Alpha adds an RGBA LoRA with separate alpha and rgb VAEs, and InfiniteTalk arrives as model patches rather than a full model.

Download the repository files
Place them in the ComfyUI model folders
Open a listed workflow
Select the model, text encoder and VAE
Run it and get a clip

Good to know:

• Many files: one repository holds many model variants, so you only need the ones your workflow names
• Original sources: the models come from Wan-AI, alibaba-pai, MAGREF-Video, htdong and MeiGen-AI
• Other versions: other people have published 4 quantized and 3 fine-tuned versions
• License: Apache 2.0
• Cost to run: none named
• Data it reaches: the files you download and the prompts, images or audio you load in ComfyUI
• Not for: use outside ComfyUI, or dubbing anyone's footage or voice without the right to use it

On Hugging Face, 13 November 2026

• Library: diffusion-single-file
• Downloads in the last 30 days: 4,888,894
• Likes: 1,044
• License: apache-2.0
• Based on: Wan-AI/Wan2.1-FLF2V-14B-720P
• Quantized versions: 4
• Fine-tunes: 3
• Tags: diffusion-single-file, comfyui

Links

• Model: https://huggingface.co/Comfy-Org/Wan-2.1
• Files: https://huggingface.co/Comfy-Org/Wan-2.1/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download Comfy-Org/Wan-2.1

Then follow the usage in its model card. Nothing in this kit runs by itself.