I see models with no .json lately, how do I use them ? (only .pth, .index, and .npy files)

[FoxTrotte]

Hey everyone !

I started using this soft a couple of weeks ago, and as I was looking around for models to try it with, I started seeing more and more models being shared without a .json, and only a .pth, .index, and .npy file. The .pth being much smaller than usual.

Are these models currently unsupported, or are their any ways to usse them with so-vits-svc-fork ?

Thank you very much !!

[gvydon]

I have the same question too. Someone, please answer.

[sonphantrung]

Please give me an example of such model. Plus, do they have a tutorial on how to run that model?

[sonphantrung]

Strange, since models should always come with a config file (heck, all so-vits-svc 4.0 derived colabs/apps require a config file to run). And AFAICT, the files are in .zip format, did you extract the downloaded .zip files to see the content inside?

[Grownz]

Can you give an example? DL'ed one as an example and it had both .pth and .json files in it:
**

**

[gvydon]

https://huggingface.co/QuickWick/Music-AI-Voices/tree/main/Dio%20Brando%20(From%20JoJo's%20Bizzare%20Adventure)%20(RVC)%2010k
https://huggingface.co/QuickWick/Music-AI-Voices/tree/main/Eminem(Slim%20Shady%20LP%20trained%20)%20800%20epoch%20Small%20Edition
https://huggingface.co/QuickWick/Music-AI-Voices/tree/main/Duke%20Nukem%20(RVC)%20300%20Epoch
There are a lot of them. Maybe there is a way to make a .json file yourself? I don't know. I've seen colabs that work with .zip files, but I am not sure if they will work with these kind of models. I think those colabs just exctract zip and use the same file structure as so-vits-svc-fork. Here's a code for one of them: https://i.imgur.com/5DzsnV0.png

[Grownz]

Ok, it seems that the combination of .pth and .index Files is for another voice changer: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI

The files are mentioned in their FAQ:

[gvydon]

Thank you, I'll try it!