>>50286249I've been using so-vits-svc-fork. I'm on Linux/AMD, but it's designed for Windows/Nvidia. It's pretty easy to get going, just follow the steps on the page to start training, or download premade models.
Vtubers are easy mode, I use yt-dlp to rip a just talking stream and Ultimate Vocal Remover to cleanup the audio and remove bgm. After that, just let it bake for 200-500 epochs.
https://github.com/voicepaw/so-vits-svc-forkhttps://github.com/Anjok07/ultimatevocalremoverguiSimilar workflow for re-dubbing a song. Just download, use ultimate vocal remover with UVR-MDX-NET Main (and a second pass using Karaoke 2 if there are backup vocals), run the vocals through so-vits-svc-fork, then recombine vocals and instrumentals in audacity. It's actually pretty easy.
One more Pippa song for fun
https://vocaroo.com/19skPOhg4DM5 