Open-Source Karaoke from any YouTube video.

Need linux or WSL2 for audio processing.

Have FFMPEG and FFPROBE installed and in your PATH.

We haven't tested on machines that do not have CUDA enabled.

The backend expects the frontend/dist folder to be built and present.

cd frontend
npm run build

Have the following in a .env file in the backend directory:

GENIUS_TOKEN=<your_genius_token_here>
GEMINI_API_KEY=<your_gemini_api_key_here>

cd backend
conda create -n okaraoke python=3.13
conda activate okaraoke
pip install -r requirements.txt
fastapi dev okaraoke/main.py 

cd frontend
npm run dev

You need to have C++ build tools installed on your system.