Character from the mobile rpg game housamo, voiced by Junko Noda (野田順子), she is also the voice actress of Veemon from Digimon. Dataset: Sample rate: 40k Language: Japanese 5:46 minutes of data, ripped from the game files No UVR/Mvsep used. Training settings: Mangio-crepe 8 Hop Length 16 Batch size 350 Epochs Normal Pretrain used Download link: https://huggingface.co/lyery/mymodels/resolve/main/zhurong.zip?download=true

talking demo

singing demo

I forgot to tell u to resample to 32k and train on dat :<

:cri: i dont remember which sample rate the audio had

Everything ripped from yt has compressed frequencies above 16khz

40k audios should be downsampled to 32k?

Ooo

Ye so it's mostly trash frequencies that don't help when labeling Audio

Don't worry it's fine though

true, welp at least i know what to do if i have a dataset from yt

Thanks