Description

download rmvpe is in the title this time no singing sample because im too lazy to find new singing audios, but the model can sing just fine (can someone share me singing audios for inference???plsplspls) info: minutes (7 minutes of real data, 7 minutes of augmented data) batch size: 8 sample rate: 32k pretrain: f0G32k.pth (the default/original pretrain) index 0.5~0.7 to reduce pretrain leakage blah blah