A full-stack machine learning application for high-performance audio processing.
This monorepo provides a beautiful, modern React-based web interface to power several high-end AI audio scripts running locally via FastAPI, including Spotify's Basic Pitch (polyphonic transcription) and Meta's Demucs (audio stem separation).
- Stem Splitting: Isolate vocals, bass, drums, and other instruments from a master track.
- Polyphonic MIDI Transcription: Convert any audio track (or isolated stems) into a high-quality MIDI file using deep learning.
- Chord Extraction: Generate accurate
.labchord charts from any audio track using Librosa chromagrams and Viterbi decoding. - Lossless Conversion: Easily bulk convert
.flacfiles into.wav.
frontend/: A modern React SPA built with Vite. It features a premium dark-mode glassmorphism UI with drag-and-drop file upload.backend/: A Python FastAPI backend that serves as the bridge to the local ML scripts.outputs/: The centralized destination directory where all processed files (MIDI, LAB, Stems, WAV) are stored to keep your original library clean.
From the root directory, install the required dependencies:
# Install root Node tools (concurrently)
npm install
# Setup the python backend
cd backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
# Return to root
cd ..Thanks to concurrently, you can launch both the React frontend and the Python backend simultaneously with a single command from the root directory:
npm startThen, simply open your browser to http://localhost:5173.