pip install spleeter
although it is always a good idea to make sure pip is current by first doing
pip install --upgrade pip
As an example, if you have a file,
Mad World.mp3 you can split it up by
spleeter separate -i "Mad Worls.mp3" -o audio_out -n filename
The first time you run it will be a tad slow as it must first download the training files into the folder
pretrained_models. Subsequent runs are much faster. Output is in
.\audio_out\Mad World\accompaniment.wav .\audio_out\Mad World\vocals.wav
I found that the easiest way to plug myself in was to run Audacity (free audio software), drop in the accompaniment and vocals as separate stereo tracks, then add a new stereo track. Plug in the headphones, hit record and sing along. Then you can mute or delete the original vocal track and voila.
Adding command line parameters allows you to separate out instrumental tracks as separate drum, piano, etc.