Description
1. General purpose, 500 epochs: 2. Fine-tuned on "The Messenger" vocals, 500+500 epochs: 3. Fine-tuned on "The Messenger" cracked vocals, 500+500+50 epochs: All three models perform best on high-pitched vocals, especially the ones with "The Messenger" vocals. These two seem to perform particularly well with high-pitched vocals, in fact you can even try to use them on female vocals without pitch-shifting. For inference, RMVPE performs good, but mangio-crepe seems to generate better raspy/cracked/fried vocals. "Dear God" was generated with models 1 and 3, "Castle of Glass" with model 2 (and a +12 pitch shift).
Comments

tsua098
🔥! I was thinking of doing something similar with the Hybrid Theory-era Vox, so cool!
Add a comment
Samples
New
Classic
1. Singing
Male
English
2. Singing
Female
English
3. Singing (Dry)
Female
English
4. Singing (High)
Female
English
5. Singing 2
Male
English
6. Singing (Dry)
Male
English
7. Singing (Dry, High)
Male
English
Pitch
More models by