create-stems

Audio stem separation using Demucs' Python API. Default model is htdemucs_6s which splits audio into 6 sources: vocals, drums, bass, other, guitar, piano.

Usage

# 6-source separation (default)
./run.sh separate --mix song.wav --out ./stems

# Extract a specific instrument by name (auto-maps to Demucs source)
./run.sh separate --mix song.wav --out ./stems --instrument vocals
./run.sh separate --mix song.wav --out ./stems --instrument oud
./run.sh separate --mix song.wav --out ./stems --instrument piano

# Raw two-stems mode (if you know the Demucs source name)
./run.sh separate --mix song.wav --out ./stems --two-stems guitar