filliptm/DeepFaceLive — reverse-engineered prompt

Build me a desktop app like DeepFaceLive for real time face swapping and face animation.

I want a simple GUI where I can pick a webcam or video file as the input, choose a pre trained face model or a still image, and then either swap the face live or animate the still image using my expressions. It should feel usable for streaming, video calls, and content creation, with low latency and decent frame rate on a modern GPU. If a supported GPU is available, use it for acceleration, but the app should still run in a more basic mode if needed.

Please include a clean first run setup, model loading, user settings saved in a userdata folder, and clear controls for starting, stopping, and tuning the result. It should work on Windows and Linux if possible. Use the existing project structure and wire up whatever is missing so I can install dependencies and run it with the main entry point. Check the setup docs or current docs online if you need to.