FacePoke

FacePoke

Table of Contents

• Introduction
• Acknowledgements
• Installation
  • Local Setup
  • Docker Deployment
• Development
• Contributing
• License

Introduction

A real-time head transformation app.

For best performance please run the app from your own machine (local or in the cloud).

Repository: GitHub - jbilcke-hf/FacePoke

You can try the demo but it is a shared space, latency may be high if there are multiple users or if you live far from the datacenter hosting the Hugging Face Space.

Live Demo: FacePoke on Hugging Face Spaces

Funding

FacePoke is just a tiny project!

There are no plans to create a proprietary/expensive cloud-only blackbox SaaS for FacePoke, so if you like it, you can buy me a coffee 🫶

<a href="https://www.buymeacoffee.com/flngr" target="_blank"><img src="https://www.buymeacoffee.com/assets/img/custom_images/orange_img.png" alt="Buy Me A Coffee" style="height: 41px !important;width: 174px !important;box-shadow: 0px 3px 2px 0px rgba(190, 190, 190, 0.5) !important;-webkit-box-shadow: 0px 3px 2px 0px rgba(190, 190, 190, 0.5) !important;" ></a>

Acknowledgements

This project is based on LivePortrait: https://arxiv.org/abs/2407.03168

It uses the face transformation routines from https://github.com/PowerHouseMan/ComfyUI-AdvancedLivePortrait

Installation

Before you install

FacePoke has only been tested in a Linux environment, using Python 3.10 and CUDA 12.4 (so a NVIDIA GPU).

Contributions are welcome to help supporting other platforms!

Local Setup

1. Make sure you have Git and Git LFS installed globally (https://git-lfs.com):

   git lfs install
   

2. Clone the repository:

   git clone https://github.com/jbilcke-hf/FacePoke.git
   cd FacePoke
   

3. Install Python dependencies:

   Using a virtual environment (Python venv) is strongly recommended.

   FacePoke has been tested with Python 3.10.

   pip3 install --upgrade -r requirements.txt
   

4. Install frontend dependencies:

   cd client
   bun install
   

5. Build the frontend:

   bun build ./src/index.tsx --outdir ../public/
   

6. Start the backend server:

   python app.py
   

7. Open http://localhost:8080 in your web browser.

Docker Deployment

1. Build the Docker image:

   docker build -t facepoke .
   

2. Run the container:

   docker run -p 8080:8080 facepoke
   

3. To deploy to Hugging Face Spaces:

   • Fork the repository on GitHub.
   • Create a new Space on Hugging Face.
   • Connect your GitHub repository to the Space.
   • Configure the Space to use the Docker runtime.

Note: by default Hugging Face runs the main branch, so if you want to push a feature branch you need to do this:

git push <space_repo> <feature_branch>:main -f

Development

The project structure is organized as follows:

• app.py: Main backend server handling WebSocket connections.
• engine.py: Core logic.
• loader.py: Initializes and loads AI models.
• client/: Frontend React application.
  • src/: TypeScript source files.
  • public/: Static assets and built files.

Increasing the framerate

I am testing various things to increase the framerate.

One project is to only transmit the modified head, instead of the whole image.

Another one is to automatically adapt to the server and network speed.

Contributing

Contributions to FacePoke are welcome! Please read our Contributing Guidelines for details on how to submit pull requests, report issues, or request features.

License

FacePoke is released under the MIT License. See the LICENSE file for details.

Please note that while the code of LivePortrait and Insightface are open-source with "no limitation for both academic and commercial usage", the model weights trained from Insightface data are available for non-commercial research purposes only.

---

Developed with ❤️ by Julian Bilcke at Hugging Face