A tool for music seperation

I found this neat tool and thought others here might find it useful.

There is a free and open-source tool called spleeter (https://github.com/deezer/spleeter) and it allows you to take any track and isolate the vocals and instruments. I’ve found it pretty useful for stripping away instrumentation on songs where I need the vocals alone.

The code is pretty straightforward and easy to run if you know a little programming Python. I’m not associated with this project at all, but I’d be happy to help anyone who’d like to give this a try.

Sound’s pretty interesting. I am scared of that conda thingy right there. The system requirements say it needs a lot of stuff i do not have yet.

Any advice?

Oh i see there is pip as an option too. Let me check that!

Well almost. Got stuck here:

spleeter separate -i "Curse Und was ist jetzt.mp3" -p spleeter:2stems -o output
INFO:spleeter:Loading audio b'Curse Und was ist jetzt.mp3' from 0.0 to 600.0
WARNING:spleeter:module 'ffmpeg' has no attribute 'probe'

Using Python 3.6.8

Do you have ffmpeg? https://www.ffmpeg.org/download.html

i tried with pip install ffmpeg and then pip install ffmpeg-python. Both did not work =/

What version of ffmpeg do you use?

ok got it working now. Had to manually install ffmpeg like described here:

Quite stunning how good this works. Even though the voice sounds a bit electronic with my first try.

Sounds interesting, are there any demos for us to listen without violating any copyright laws?

I’ve used it well on a recording of my friend playing banjo and singing. I ran the program on their example and here’s the output: https://soundcloud.com/yakczar/sets/spleeter-example

You can tell the vocals are a little off, but its not too bad, especially if you have no alternative or if youre doing something lofi-y