Character Animation for Agents

Character Animation for Agents

Rigs a static 3D model and animates it from a one-line prompt, returning a rigged, animated GLB. Built for AI agents and developers, through REST, MCP or the browser, on Hugging Face ZeroGPU.

Made by townbox 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 is a Gradio Space that rigs and animates a 3D model from a sentence. You send a static model file (glb, gltf, obj or fbx) with a short prompt and get back a rigged, animated GLB. It chains two released research models, SkinTokens for the skeleton and skin weights, and UniMate for motion made from text. Agents can call it over REST or MCP.

Who it suits: Developers building agents or tools that need animated 3D characters, artists who want a quick motion draft, and researchers who want to try both models together.

Industry | Who uses it | How they use it | Example
Games | Indie developers | Get a draft motion for a static character | Ask for an object that walks forward
Software | Agent builders | Call the service from an agent over MCP or REST | Let an agent animate a mesh from a one-line task
3D art | Artists | Try motion ideas before animating by hand | Ask for a jump on a prepared asset
Research | Graphics researchers | Try SkinTokens and UniMate in one flow | Compare several samples for one prompt

Skip it if your project is commercial, since the license named is cc-by-nc-4, a non-commercial one.

Where to use it:

• A web browser, on the Space's page at Hugging Face
• MCP clients such as Claude Code, through the Space's MCP endpoint
• Any program that can send REST calls, which the Space documents with OpenAPI
• The Gradio Python and JavaScript clients

Use case | Who | What they get back
Animate a static model | Developers and artists | A rigged, animated GLB with 60 frames of motion, 2 s at 30 fps
Rig a model only | Pipeline builders | A predicted skeleton and skin weights for the mesh
Refine a motion | Animators | In-betweening, joint editing or multi-prompt expansion

How to set it up:

• Nothing to install for the online page
• A Hugging Face account and token, to call the service from an agent
• Python 3.12.12 and pip, only for a local copy
• The Python packages listed in requirements.txt, for a local copy

1. Open the Space's page at Hugging Face to try it online.
2. To use it from an agent, create a Hugging Face token. GPU time is charged to the caller's own account through it.
3. Add the Space as an MCP server in your client, or call its REST endpoints. The landing page offers a copyable client config and a one-line Claude Code command.
4. For a local copy, install the Hugging Face hub package and download the Space.
5. Move into the new folder, install Gradio at the pinned version and the packages, then start the app.

pip install -U huggingface_hub
hf download townbox/character-animation-for-agents --repo-type space --local-dir character-animation-for-agents
cd character-animation-for-agents
pip install gradio==6.17.3
pip install -r requirements.txt
python app.py

The repository is large, at 302 files, and the hosted Space is where the GPU steps are meant to run.

How to use it: In a browser, use the landing page. For an agent, the page has a card with a one-sentence prompt to hand over and a card with MCP connection details. One call does everything and takes about 40 to 90 seconds. Each stage also has its own call, and assets are reusable by id.

python app.py

Tool | What it does
animate_mesh | Rigs and animates a mesh in one call
rig | Predicts the skeleton and skin weights
prepare_asset | Turns a rig into a reusable asset
animate | Makes motion from a text prompt for an asset
inbetween | Fills in motion between poses
edit_joints | Edits the joints of a motion
expand | Expands a motion from several prompts
view_rig, view_animation, view_asset | Previews of the rig, the motion and the asset
service_info and guide | Information and an agent guide

Example: You tell your agent to animate a robot model with a prompt asking for an object that walks forward. The agent sends the file, the service rigs it, prepares the asset and generates 60 frames of motion. An animated GLB comes back after roughly a minute.

For newcomers and experts: If you are new, open the Space page and look at its landing cards. One gives a short prompt you can paste to an AI agent, and bundled examples let you try it without your own model. Start with one example and one short prompt.

If you know the field, the rig is predicted by TokenRig, then a preprocessing step inside the Space handles Blender import, joint labels, a facing pair and a canonical T-pose. UniMate then makes the motion. Only the TokenRig pass and the flow rollout use the GPU, and agents can correct labels, facing and rest rotation on a prepared asset.

Send a 3D model and a sentence
TokenRig predicts skeleton and skin weights
The rig becomes a reusable asset
UniMate makes 60 frames of motion
You get a rigged, animated GLB

Good to know:

• License: cc-by-nc-4
• Cost to run: no payment is named, but a Hugging Face token is used and the caller's ZeroGPU quota pays for GPU time
• Data it reaches: the 3D files you send, up to 60 MB, and results the Space keeps for 24 hours
• Not for: commercial work under this non-commercial license, or files above 60 MB

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: cc-by-nc-4.0

Links

• Space: https://huggingface.co/spaces/townbox/character-animation-for-agents
• Files: https://huggingface.co/spaces/townbox/character-animation-for-agents/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.