All Tools
O
OtherFreeOpen Source
OPEN-UNMIX
PyTorch music source separation into vocals, drums, bass, and other
MIT
ABOUT
Music and speech pipelines need a reproducible stem separator, not a closed commercial unmixer. Open-Unmix ships pretrained PyTorch models that mask spectrograms into vocals, drums, bass, and other, plus a speech enhancement checkpoint, so researchers and audio engineers evaluate and fine-tune source separation against MUSDB18.
INSTALL
pip install openunmixINTEGRATION GUIDE
1. Separate a stereo mix into vocals, drums, bass, and accompaniment stems
2. Benchmark music source separation models on MUSDB18 with museval
3. Enhance noisy speech with the umxse checkpoint before ASR
4. Train a custom Open-Unmix target model on in-house multitrack data
TAGS
pythonpytorchaudiomusicsource-separationspeech-enhancementopen-source