Spleeter to Separate Musin on Sound and Vocals

AI learned to distinguish between vocals and melody in songs. The tool is called Spleeter, it is the internal technology of the streaming service Deezer.

Spleeter can divide a track into several components. There can be a maximum of five: bass, drums, piano, vocals and the rest of the background. The user downloads audio, and the output receives several files.

Spleeter is powered by TensorFlow. According to the developers, the model was trained on tens of thousands of songs. While it works intermittently, it leaves artifacts on the tracks.

The neural network library is on GitHub and is distributed under the MIT license. True, to use Spleeter, you need to fumble in Python and TensorFlow minimally. In addition, interaction takes place through the command line, there is really no interface yet.