faint field
comeback with Indonesian Artist again huh (or vtuber)
unlike the @gilded falcon model that focusing on speech data ig, this model based from her singing on most of original song and featured. if you don't familiar check out the @gilded falcon moona singing model
actually i wanna use 40k sample rate but somehow i only got 17k worthy dataset from her vocal so 32k is the only solution
Source : any song from Kobo Kanaeru
Duration : 11 minutes 35 seconds
RVC version : V2
Pitch extraction : RMVPE
Pretrain : KLM 4 32k Hifigan
Batch Size : 4
Epochs : 310
Steps : 20,150
model and dataset by @faint field
https://huggingface.co/ABF54/ABFRVC/resolve/main/KoboKanaeruSingingModel.zip?download=true
mighty pewter