Description

(CREDIT ME IF USED) "m19 [kei]", russian artist on Spotify, Deezer, and more. Check her out, her music is heavenly and i spent a lot of time on this 48k model of her cuz i loved her voice that much. Talented as fuck. Please do russian samples for me since I dont have any russian songs isolated yet. i got , Kino, and even models planned, theyll be 48k too (if i can do it, this model was so difficult to make completely without UVR) Dataset was 16 minutes of silence truncated and curated data by me, completely without UVR, 44.1k vocals isolated with demucs on CLI. THIS MODEL WORKS BEST AT LIKE 0.7 SEARCH RATIO. the echo was seriously difficult without UVR but im getting better at deechoing manually, just practicing again and again and i really am getting better at it.