Deezer, a music streaming service provider, has released an open-source tool on Github that uses machine learning to split a finished track into drums, vocals, bass, and others. Spleeter comes with pre-trained models for 2, 4, and 5 track separation. The splitting process is a lot faster than real-time, although it’s not perfect but impressive.
Spleeter comes in the form of a python library based on TensorFlow, and it is available on Github.
“Spleeter will be presented and live-demoed at the 2019 ISMIR conference in Delft.”: Deezer
Github: https://github.com/deezer/spleeter
Details: https://deezer.io/releasing-spleeter-deezer-r-d-source-separation-engine-2b88985e797e
