MoCapAnything V2

MoCapAnything V2

Turns a video into a predicted 3D pose skeleton for any species or object, with BVH and npy files to download. Retargets motion onto a skeleton that differs from the one in the video.

Made by kehong 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: MoCapAnything V2 is a motion capture tool for arbitrary skeletons. You pick a video and a target species or object, and it predicts a 3D pose skeleton. The target can differ from what is in the video, which is called retargeting. You can then download the motion as BVH and npy files.

Who it suits: Animators, game developers and computer vision researchers who need motion data from video, including for creatures that differ from the one filmed.

Industry | Who uses it | How they use it | Example
Animation | Animators | Turn a video into a skeleton motion they can load into animation tools | Download the BVH file for a clip
Games | Game developers | Retarget motion onto a skeleton of another species | Pick a different species as the target
Research | Computer vision researchers | Test motion capture on skeletons of many kinds | Run the demo and read the npy output
Education | Students | See how video becomes 3D pose data | Compare the input video with the skeleton result

Anyone who needs a mesh render or a finished animation from this Space alone can skip it, since it shows the skeleton output.

Where to use it: It runs in these places.

• A web browser, on the Space's page at Hugging Face
• Windows, with Python
• macOS, with Python
• Linux, with Python

Use case | Who | What they get back
Capture motion from a video | Animators | A predicted 3D pose skeleton
Retarget to another skeleton | Game developers | Motion for a different species or object
Export the motion | Researchers and animators | BVH and npy downloads

How to set it up: Using the Space online needs only a browser. A copy on your own computer needs the following.

• Python and pip
• The huggingface_hub package, which gives you the hf command
• Disk space for the repository, which holds 4 files

1. Install the huggingface_hub package.
2. Download the Space's files into a local folder.
3. Move into that folder.
4. Install Gradio at the version the Space uses, then install the packages in requirements.txt.
5. Start the app.

pip install -U huggingface_hub
hf download kehong/MoCapAnythingV2 --repo-type space --local-dir MoCapAnythingV2
cd MoCapAnythingV2
pip install gradio==6.20.0
pip install -r requirements.txt
python app.py

How to use it: Open the Space's page, or run the app on your computer and open the page Gradio shows. Pick a video, pick the target species or object, and run it.

python app.py

Tool | What it does
Video | The input clip that holds the motion
Target species or object | The skeleton to predict, which may differ from the input
Skeleton output | Shows the predicted 3D pose skeleton
BVH and npy downloads | Files holding the predicted motion

Example: You pick a video of a person walking and choose a different species as the target. You get back a 3D pose skeleton for that target, plus BVH and npy files to download.

For newcomers and experts: Motion capture records how a body moves. Here the recording is a normal video, and the result is a skeleton that moves the same way. The first small thing to try is the online Space with a short clip.

The project names a weights repository and a sample data set. The Space shows the skeleton output only. The local demo in the project's code repository adds a Blender 3D mesh render and a Dance Anything tab, where a dance video and music make an animal perform the dance. The Space is set to run on Hugging Face's zero-a10g hardware.

You pick a video
You pick a target species or object
The model predicts a 3D pose skeleton
You download BVH and npy files

Good to know: The extra tabs belong to the local demo, so the Space shows less than the full project.

• License: mit
• Cost to run: none named
• Data it reaches: the video you pick, plus the weights and sample data that Hugging Face supplies
• Not for: the Blender mesh render and the Dance Anything tab, which only the local demo offers
• Hardware: the Space is set to zero-a10g, and what your own computer needs is not stated

On Hugging Face, 13 November 2026

• Likes: 16
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/kehong/MoCapAnythingV2
• Files: https://huggingface.co/spaces/kehong/MoCapAnythingV2/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.