Wan 2.2 Animate Character Replace

Wan 2.2 Animate Character Replace

Replaces the character in a video with one from a reference image, using the Wan 2.2 Animate model on ZeroGPU. It suits video creators and animators who want to swap a character without a local setup.

Made by neksjgg on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space swaps the character in a source video for a character from a reference image. Detection models ViTPose, YOLO and SAM2 pick out the person's pose, face and silhouette in your clip. Wan 2.2 Animate then draws the clip again with the new character. It runs on zero-a10g hardware at Hugging Face.

Who it suits: It suits people who edit or animate video and want to try a character swap, and developers who want to see a ComfyUI workflow run behind a Gradio page.

Industry | Who uses it | How they use it | Example
Video production | Editors and creators | Swap a character in a short clip | Put a designed character into a recorded performance
Animation | Animators | Test a character on existing motion | Move a new character with the motion of a source video
AI development | Developers | Study a ComfyUI workflow wrapped in Gradio | Read app.py to see how the exported workflow is driven

You can skip it if you need long renders, since a free ZeroGPU account limits how long a run can take.

Where to use it: You can use it in these places:

• A browser: open the Space's page on Hugging Face and nothing needs installing.
• Windows, macOS or Linux: run your own copy with Python and a GPU with plenty of memory.

Use case | Who | What they get back
Character replacement | Video creators | A video re-rendered with your reference character
Motion transfer | Animators | The source video's pose and movement on a new character

How to set it up: You need these first for a local copy:

• Python and pip
• A GPU with a lot of memory, since the hosted run gets 48 GB by default
• Disk space and time for the first launch, which downloads about 20 to 30 GB of models

The online page needs none of this. For your own copy, follow the steps below.

1. Install the Hugging Face tools.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio at the version the Space uses.
5. Start the app.

pip install -U huggingface_hub
hf download neksjgg/wan22-animate-character-replace --repo-type space --local-dir wan22-animate-character-replace
cd wan22-animate-character-replace
pip install gradio==5.43.1
python app.py

How to use it: Open the Space in a browser, or start your own copy with the command below. Give it a picture of the character you want and the video to change, then start the run.

python app.py

Tool | What it does
Reference image | The character that gets inserted into the video
Source video | The video whose character is replaced
Frame Load Cap | A setting to keep low while testing
Detection models | ViTPose, YOLO and SAM2 find where the person is and how they move

Example: You upload a picture of a character and a short source video. The Space finds the person's pose and mask in the video and returns the video again with your character in their place.

For newcomers and experts: If you are new to this, use a short clip with a low Frame Load Cap on your first run. Free ZeroGPU time is limited, so a short test saves your quota.

If you know the field, the Space is a vendored ComfyUI checkout with four custom node packs, driven without a UI by app.py. Models that never change are loaded a single time when the app starts. Only the steps that depend on your input run on the GPU for each request. The duration is set to 120 seconds, because the free tier rejects longer requests.

Upload an image and a video
Pose, face and mask are detected
Wan 2.2 Animate renders the video
You get the swapped video

Good to know: A free account has a daily ZeroGPU quota of five minutes in total, so a few long runs can use it all. If you hit out-of-memory errors, the README suggests asking for the xlarge GPU size, which has 96 GB and costs twice the quota. The first launch is slow because it downloads the models.

• License: apache-2
• Cost to run: ZeroGPU quota, which is five minutes a day on a free account, with a PRO account giving more
• Data it reaches: the image and video you upload, and model files downloaded from the Hub on first launch
• Not for: long renders on a free ZeroGPU account

On Hugging Face, 13 November 2026

• Likes: 10
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/neksjgg/wan22-animate-character-replace
• Files: https://huggingface.co/spaces/neksjgg/wan22-animate-character-replace/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.