nekalakininahappenenawiwanatin

makalakinevahappenamewewerenothin

your dataset you've trained has some noisy background 😅

i hope there were clean acapella of his voice

no need to send that. but i suggest them to deverb noisy sounds

that's what i used it before training on my own

ok. i might try that if i have some time

If you're gonna make a V2, consider using filtered vocals from the rest of the album: https://open.spotify.com/album/1HjSyGjmLNjRAKgT9t1cna

(also listen to it because it's good)

my fffucking ears

That model was meant to be a meme, not something well-executed

I'm gonna try and do a legit Gotye model

On a side note, is there any way to reliably separate vocal doubling?

idk

if it's not 2 vocal takes paned 100% to the left and 100% to the right (like in this for example https://youtu.be/G8IuFLRPN9Y?t=6 ), it's currently impossible

But you didn’t have to AAAAAAAAAH