No it is fully offline. It just downloads on the first instance of setup. Just like you need to download the package to install from the internet .

as for models the link to I put in my post to www.weights.gg has over 16k models to use or blend with other models and make an infinite combination of custom AI voices.

 

and

Thats the next feature we will be adding. For now you can train for free using.


Google Colab: https://colab.research.google.com/github/ddPn08/rvc-webui-colab/blob/main/rvc-webui.ipynb

 

Last edited: Nov 8, 2023

Thanks for your response.
I understand that the program works offline once it gets required files. I was asking you to provide direct links to these files, so that I could download them on a different computer and move them to my offline machine to appropriate place(s), that I also asked you to point out.

 

This would be huge! Can't wait to try it. BTW, will this feature require GPU to work?

 

great program mate
Im using v1.81

thanks a lot

 

I appreciate you! We update it frequently so keep updates on!

 

Yes GPU is best for faster rendering

Mac arm M1 and above MPS GPU works
Same with PC.
Mac Intel can only use CPU at this date.Its just slow is all

 

its in the orig post but here is a link to all the 16k+ models

www.weights.gg

 

I think there's a major misunderstanding here. My DAW machine is offline, so I'm not willing to
let neither yours, nor any other program connect to the internet. Therefore in order to use Replay I need direct links to the 18 files it tries to download at startup, so that i can download them manually, as well the information where to put them on the offline machine, once downloaded.

In case you can't or don't want to provide these links, please consider removing the requirement of presence of obligatory default voice(s) from the program, and let users decide which voices they want use and let them choose the method of acquiring the voices.

 

No worries! The use case for our app isn't for you!

The initial models are actual models to run the app. The way the algo works is it needs the initial models to even run. They are the guidelines of how the human voice is comprised.

A lot of these AI tools need trained models to even run.

Then the user can put in the models they want that refer to the default dataset models that are the scaffolding for the python algo.

 

Well, it could be for me (and several other offline users, I guess) if you'd either:

1. Provide direct links to these initial models, so that I can download them manually and put them in appropriate place(s). No idea why am I being forced to let the app itself download them.

or

2. Incoroprate the initial models into the app installer. I does not necessarily mean 18 of them. I think one or two will do to try the program.

PS.
Are these 18 files all voice models, or are there also other types of files among them?