Okay, later, if you want, I can share what I have, but I think you have a different format. That is, my Colab saves a file that says 'Weights' and the other one '.pts' (I don't remember the extension well because it's always saved in .zip). Do you think they are compatible? Can your models be used in Colab? Because the software is impossible for me, right now I'm precisely cloning a voice in Colab, and it's going quite fast.

 

Bon scot AI - Feat - Thunder Mother - It´s Just a Tease

 

the file format is .pth and yes replay accepts that file format

 

he's unstoppable!

 

Steve Lee AI -Gotthard Feat - Thundermother- You Cant Handle Me
Trained Pack is @https://www.weights.gg/

 

Well, I've downloaded a model, and it has three files: model.pth, model.metadata.json, and model.index. I always place the .pth file in the 'weight' folder along with another file that is something like 'added_IVF1268_Flat_', which goes inside the 'Logs' folder. Then, in the folder of the loaded model, in this case, that file (added_IVF1268_Flat_nprobe_1) is not present in your models. As for the other two files mentioned earlier, I wouldn't know where to put them. Let's see if you can help me figure out how to test your models. On the other hand, I noticed you only have an electric bass; I have one that sounds quite decent, also a saxophone, and others. However, we would face the same dilemma since I would have other configuration files.

 

Theres no reason to unzip these files Leave the index file and the .pth in the same zip File
With replay launched on your screen drag and drop the zip file onto the custom rvc model to be used area on the screen and replay will automatically install the model for you to use,
the model you install will then be in the menu of trained models for future use.

 

Last edited: Dec 4, 2023

It's not for use in the program; it's for use in Colab. This conversion comes from another one, where I told him that I would like to try his models in the Colab I use for cloning instruments since my PC doesn't support this program. That is, it supports it terribly slowly, taking up all my CPU and RAM. So, if you read above, I asked him, 'Do you think it will work in Colab?' and he said yes, that the files were the same, so it should work, but they have different files.

 

oh i see my apologies, I have not used the colab unfortunately so I cannot give comment on how to use a colab or the process's needed to have a colab function my friend
i'm sorry i cannot be more helpful with colab.

 

Last edited: Dec 4, 2023

No worries, what matters is the intention. Thank you for wanting to help

 

Trained
Joe Elliot - Def leppard
https://www.weights.gg/

 

what colab are you using? What version of RVC is it using. Because it should be making .pth files. Is it not?

 

My initial download of the required files keeps just getting stuck randomly. No issues with my internet. THe download just stops in the middle of a progress bar. It's not the same spot each time.

EDIT: Nevermind - 4th time's the charm.

 

Last edited: Dec 4, 2023

Good day curtified, it's one of the first Colabs that was released. I had to modify it because the code is banned. I can share the Colab with you, but it might be a bit overwhelming because I added 'Tools' like downloads, zip, converting audio from wav to flac, etc. My craziness understands it, haha. This Colab was initially created by 'Retrieval-based-Voice-Conversion-WebUI,' surely it's RVC 1, and it has the PTH file, but it also has another one that I mentioned before. Your Zip contains three files. When I try to load the model into the RVC, it gives an error

 

After a few days of fun this will be the last one for a while work is calling
Recording Engineering to do
this program has really been alot of fun once you get the flow.

Wendy James AI- Transvision - Kids In America

Brian Johnson ACDC feat Thunder Mother - We Fight in the name of Rock N Roll

 

Last edited: Dec 5, 2023

T

thanks for the information. I tested replicate with 50% success. I managed to model Sting from an interview available on Youtube. OK it's not bad.
Using the same method, I tried to model Stevie Wonder and failed (pw error) after an hour of trying to convert.
I successfully modeled my own voice and my woman by reading a text, singing A B C D E F G A on 2 octave. It's ok.
I used replay to load my voice onto a track and the result was so good that my father said it sounded like your voice but in disguise. My wife also recognized my tone.

is there a free alternative with a graphical interface to replicate?

Also, I'd like to understand why I'm having these failures with replicate.
Dossier dataset OK, 48K stereo OK, audio fragments of +or- 10 seconds clear at -12db,
approx. 30 to 50 Split_0, Split_1 up to the last piece for a duration of approx. 10mn. zip format of the whole set.

2 echec cout sur cout s'en comprend ce que j'ai fait de mal?

Are there any hidden audio protections on Youtube?

 

Seems version 3.0 is out so whats New ?
a update log would be nice

 

hi, how do I turn off the download. every time I open the App it wants to download...


this time with internet off it worked.

 

Last edited: Dec 7, 2023

I love this App! :-D
Just the resolution.It is only 16 bit, right?

 

Can someone give instructions om how to run this app through Google Colab? Thanks