if you're going to use this model, id recommend experimenting with the different methods (crepe, harvest, dio, etc) to create the songs, because some of them need to be layered together to sound like anything good

also experiment with layering the normal vocals in crepe with a tune down of -12

r

Looks like I found the one without the fetish image.

Context (A later RVMPE version modified the image to have a shocked image and two arrows pointing at the characters feet for... some reason...)

lol just checked it out... it sounds like the same dataset was used haha, just trained on a different version

both of our models have that grouchy sound tendency