We present and release a new tool for music source separation with pre-trained models called Spleeter. It was designed with ease of use, separation performance and speed in mind. Spleeter was trained and distributed with Tensorflow framework and makes it possible both:
- Separate audio files into 2, 4 or 5 stems with a single command line using pre-trained models.
- Train source separation models or fine-tune pre-trained ones with Tensorflow (provided you have a dataset of isolated sources).
We provide multiple ways of running it, either as easy to install Python package from PyPi, or a self contained solution through Conda or Docker. As we released Spleeter publicly we also submitted an official paper to the Journal of Open Source Software which was accepted and published on which you can dive into implementation and distribution details.
