Description
1. All rounder model, mangio-crepe, 250 epochs: 2. Raspier / screaming model, mangio-crepe, fine tuned on the all rounder, 250 + 50 epochs: 3. Quiet singing / whispering, fine tuned on the all rounder, 250 (mangio-crepe) + 100 (rvmpe) epochs: 4. Quiet singing / whispering, rvmpe, 300 epochs: The provided example uses models 1 and 2
Comments

playing around with songs and models, and holy shit.. serj absolutely crushes this opera song. like, im tempted to call the cops and report a murder, because there is nothing left of the body. absolutely amazing model, and thank you for sharing it. it's really really good. (ps boats and hoes)

helped? nah, all credit due, the song was absolutely *made* by your model. super graceful. opera is hard for a lot of models since inflection is so nuanced, they need to be very well balanced with their dataset. my alex model is a little too biased towards yelling (which is by design because it's funnier), but your main serj model is extremely well balanced, the build-up was *so* graceful and it masterfully transitioned into the power notes. I can't think of many (if any) that could handle that so well *and* make it sound so natural. Extremely well executed. bravo.

Baby underrated af song.
Add a comment
Samples
Pitch
More models by
More to explore
Loading more