Giters

Mangio621 / Mangio-RVC-Fork

*CREPE+HYBRID TRAINING* A very experimental fork of the Retrieval-based-Voice-Conversion-WebUI repo that incorporates a variety of other f0 methods, along with a hybrid f0 nanmedian method.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Pretty much doesn't work anymore - most likely abandoned project

SheMelody opened this issue · comments

commented

This project is most likely abandoned, you shouldn't attempt to make this work unless you know what you're doing. You can get this to infer after editing python files, with existing models of course, but it's not worth your time anyway. Moreover, you can't train anymore using this unless you manually install old dependencies. Any trained model generated with this, using the default requirements files, will just be useless as the models are borked and unusable.

commented

A few workarounds if you really want this to work.

  • Replace all references to FFmpeg and FFprobe in scripts to ffmpeg and ffprobe (both lowercase, that's important).
  • Downgrade packages that are causing random errors to earlier versions, I can't recall the exact ones but you'll notice.
  • If using Windows, don't ever use backslashes in training paths, replace them with regular slashes (/)
commented

Well - I can add to that...
I've actually have a running docker-image, which I use to train models and do voice-cloning with them (say by either have my voice processed fo projects like audiobooks or do it live for some PnP-Online-Sessions).
If you wanna use it: https://hub.docker.com/r/cherrymint/mangio_rvc
I do NOT give support and I don't know how I will keep it up.

commented

Well - I can add to that... I've actually have a running docker-image, which I use to train models and do voice-cloning with them (say by either have my voice processed fo projects like audiobooks or do it live for some PnP-Online-Sessions). If you wanna use it: https://hub.docker.com/r/cherrymint/mangio_rvc I do NOT give support and I don't know how I will keep it up.

I'm currently using this: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI
It's better than this fork, even the produced audio has almost no artifacts with models trained for just a few minutes.
I personally don't like docker anyway.

commented

@SheMelody - There seems to be some drama goin on at RVC-WebUI... I'm resetting my docker-hub-containers to https://hub.docker.com/r/cherrymint/rvc_webui where I will provide the magio-tag for this repo.