sample :

Credit me if you want to use this so i can see what song that my AI model can sing https://www.youtube.com/channel/UCn5gYeQ0NDM4l3MNUEIiPoQ

youtube : Myura

kira kira doki doki model (do yukina <:LFG:1093368528637276191> )

Kasumi cover of Argonavis - Goal Line (pitch +4)

the voice breaks a lot unfortunately, but this is a common problem I get when making female characters sing male songs, so it might not be specific to this model

yeahh, and its because i set the pitch extraction to 128 when training, im gonna decrease it later

its on my pc, i send to you later

how's about my dataset??

well this is my cleanest dataset that i can made, hope you like it

Status : -Currently i Re-train this model and Reaching 200 Epoch, maybe my target is 300 or same (250 Epoch)

-I'm also decrease the hop to 32

Toyama Kasumi (BanG Dream!) (RVC V2) 250 Epoch - 32 Hop-Length

sample : https://youtu.be/ENnIC8iHxp4 (so far this is the best model i ever made)

i'll do yukina later

Toyama Kasumi (BanG Dream!) (RVC V2) 500 Epoch

Updated Link (use this one) https://huggingface.co/Icchan/Myura/resolve/main/ToyamaKasumiV3.zip -more Epoch -balancing hop-length to 64 (default) i suggest you to using default inference settings, it sounds good

I used your model for a cover https://www.youtube.com/watch?v=TXTkykl54tY&ab_channel=SaintA-RISEPassion