Sounds interesting. I've used zynaptiq's Stem Maker included in Acid Pro 10, and for some source material (like hip-hop that pretty much has only voice, bass, drums, and some stuff on the side) it's strikingly effective. For a typical multi-layered production...not so much .There are some demos on the Acid Pro landing page.
I've also used iZotope's Rebalancer, which alters levels for drums, bass, and/or voice within a mix. The performance is sort of like pitch transposition - the greater the amount of change, the more artifacts you bring into the picture. I've never had to actually use it in a production, because it's easy enough to tweak the levels in a project with automation, and run off another mix.
But as the Spleeter people point out, AI is making a huge difference in improving the algorithms. The difference between Stem Maker 1 and Stem Maker 2 is considerable. I'm also intrigued by the Sonnox drum gate that lets you gate specific drums (e.g., to get rid of leakage). It seems a "drum separator" for a two-track drum mix into individual drums wouldn't be too far away.
Then again, it would probably sound better just to hire you to play the part, and record what you do into a multitrack recorder