please don't take down for poor quality!
trained on a 0:19-second dataset ripped directly from YT. (mel-roformer, rx de-ess, mouth de-click, resample + eq, renegate noise-gate) ||tip: export your dataset in wav format (32-bit) instead of flac to prevent spectogram issues||
pitch extraction: rmvpe
steps: 2.7k
batch size: 2
pretrain: original v2 / 32k
precision: fp32
cv. Jakubov Brloh
please don't forget to credit me when you use this model.