Demucs is an AI software for music source separation developed by facebook research. You give it an audio file – i.e. a song track – and, after some processing, it will output as result four audio files:

1. Drums
2. Bass
3. Vocals
4. Other instruments

The official github project page has instructions on how to install the required software and execute it locally. It's really straightforward. With a bit of technical knowledge and patiance: boom! You get your tracks separted.

The problem

It can be a bit daunting for someone with no technical background. Even for someone with heaps of experience, it could get cumbersome if you find yourself executing the separation process often. Even if you are doing it just once in a while, you whether have to keep the environment setup and remember how to proceed or just setup the environment all over again if you switch machines or just wiped it out.

The service I'd like to use

Wouldn't it be nice if there was a service where you can upload a music track and then get the tracks back? Well, I thought so! And then I've been working on a 'Demucs App' side project for the last few months :D

First public beta!

I'm releasing today the first public version! It features:

1. Create a music source separation from:

   1. a mp3 track
   2. a youtube music clip link
2. Email notification – just leave an email address and you'll get the tracks on your inbox as soon as the process is done.

It still needs tons of work and polishing but I thinkg it's in a 'good enough' shape to start trying it out.

You can find it here: https://demucs.devnano.me And, also, you can check out a 'how it works' video below:

Demucs Web Beta 1 Demo