All Tools
A
OtherFreeOpen Source
ASTEROID
PyTorch toolkit for audio source separation research
MIT
ABOUT
Speech and music separation papers ship one-off training code. Asteroid puts Conv-TasNet-style models, datasets, and SI-SDR metrics behind one PyTorch API so researchers train and compare separators without rewriting each recipe.
INSTALL
pip install asteroidINTEGRATION GUIDE
1. Train speech separation models on WHAM or LibriMix recipes
2. Evaluate source separators with SI-SDR and related metrics
3. Run music or speech enhancement inference from pretrained Asteroid models
4. Prototype new separator architectures on shared Asteroid datasets
TAGS
pythonaudiosource-separationspeechpytorchmusicdsp