first model trained on ilaria rvc mainline colab 🤯 trained on 13:30 minutes of her speaking from nekopara vol. 4 cv. Tezuka Ryouko pitch extraction: rmvpe steps: 19.4k steps(?) batch size: 7 pretrain: regular v2 / 40k huggingface: https://huggingface.co/sxndypz/rvc-v2-models/resolve/main/coconutv2.zip

samples idk

You trained this using Ilaria Mainline beta?

I think this model is more suitable for japanese audio than english

yes

colab

next model on my list is saya

also from phase connect

Sakana next plz

as for sakana, i'll think about it

just have to find the clean samples from him

like that stream from jelly 3 months ago

his mic is clean enough imo

In case of anything, you can use RX10 or BS Roformer for deleting mic bleed.

that's what i use

great news, i've made the sakana model. #H2oSakana (CEO of Phase Connect) (RVC v2) - 520 Epochs