x-minus.pro/ai
Almost entirely no vocal here. X-Minus neural vocal separation rules.
Though using one algorithm is often not enough: you should experiment and use the best parts of several algorithms. For example, I used MDX regular (to remove all vocals), MDX back vocal (to restore a muffled guitar part), and UVR-post-processed ("☆ Enhance" button) versions of both of those (to get rid of some vocal bleeding).
And don't forget to check vocal tracks for some instrument leftovers - AI is NOT perfect. That "eeeEEEEEEEEH" couldn't be removed completely, but the rest of the song is now karaoke.
30 сен 2024