Cool model but I think it could’ve been improved by a long shot. It sounds like you took the vocals from isolating ai which could be good but it sounds like something I would hear isolated 2 years ago. If you can I strongly recommend either studio stems or the next best thing RockBand stems. I also think it could’ve been overtrained but I wouldn’t really know much about that my limit is 1K. Love to see this model improve there’s potential 100%

The model uses stems from Best of You The Pretender and one other song. The stems are simply not the highest quality.

Could I see a sample?

from the stems?

isn't that just copyright?

Just the Vox

edited it to remove all of the harmony sections

because the deharmonizer lowers the audio quality too much