Mizuki Akiyama (Project Sekai) (RVC v2) - 800 Epochs | AI HUB | Page 1

acoustic skiff Nov 2, 2025, 3:39 AM

#

technically the first project sekai model i've made in a while

this character is featured in project sekai.

trained on 1:20 minutes of audio data from the "a miku who can't sing" movie
(5.1 + revive 3e + anvuew karaoke (in some parts), resampled to 40k)

CV: Hinata Sato

pitch extraction: rmvpe
steps: 13.6k
batch size: 2
pretrain: klm 4 hfg / 40k

huggingface: https://huggingface.co/sxndypz/rvc-v2-models/resolve/main/mizuki.zip

#

lyric coralBOT Nov 2, 2025, 4:04 AM

#

Model Ready ✅

This model has been synced with Weights and is ready to use for free!

lyric coralBOT Nov 2, 2025, 4:04 AM

#

lyric coral

acoustic skiff Nov 2, 2025, 10:32 AM

#

Mizuki Akiyama (Project Sekai) (RVC v2) - 800 Epochs

#

it took me THIIIIIS long to notice a typo in the name 😭

blazing swallow Nov 2, 2025, 1:41 PM

#

how to use revive 3e?

acoustic skiff Nov 2, 2025, 1:52 PM

#

you install the ckpt and yaml from this hf space (https://huggingface.co/pcunwa/BS-Roformer-Revive) and put them into uvr5 (ckpt should be in MDX-Net Models, yaml should be in "MDX-Net Models/model_data/mdx_c_configs")

pcunwa/BS-Roformer-Revive · Hugging Face

#

if you don't wanna do all that you can use the msst colab:

https://colab.research.google.com/github/jarredou/Music-Source-Separation-Training-Colab-Inference/blob/main/Music_Source_Separation_Training_(Colab_Inference).ipynb

Google Colab

blazing swallow Nov 6, 2025, 12:59 AM

#

wym by 5.1? does it stand for audio quality?

acoustic skiff Nov 6, 2025, 4:21 AM

#

it means that the source for the dataset i used has 5.1 audio

#Mizuki Akiyama (Project Sekai) (RVC v2) - 800 Epochs