#11labs

A Voice Model

First, you need to choose a voice you would like to use. You can find a voice model on weights.gg or on the Discord channel

You can train your own voice model, if you didn't find the one you need or if the existing one is not good for you. The only working Notebook on Google Colab to train your model is RVC v2 Disconnected Here is the guide for it.

(I still have no idea of how to restart the training if something went wrong and the process stopped/disconnected, sorry. Also, remember that Google will ask you if you are still there and will give you a captcha to solve in a few hours, and if you'll not react to it quickly, the process will be stopped, so don't go too far from your PC and check on the process from time to time).

a tip to remember: if your dataset audio is under 10 min = 100-200 epochs. Over 10 min = 200-300 epochs. I do 500-600 if my audio is 15+ minutes, but it can overtrain the model.

Ok, you have your model, now you need to decide what you want to do with it.

If you want to create a MESSAGE:

You need Mangio-RVC-Fork installed on your PC.

2. Crate your text. Write a text with a pronounced accent of your character like:

You can ask ChatGPT to became your redactor, correct all the mistakes and add a thick accent. (For example, a thick Southern drawl for Graves, a thick Scottish accent for Soap, a thick Manchester accent for Ghost etc.)

3. Then you go to 11Labs and generate the body of your message with their models. I use Clyde for both Graves and Soap (yes, he is American, but we already have a text with accent so it's ok). 11Labs has a lot of different models, so you can listen to them and choose the one you think would be the best. Then you just generate the message. Don't put all the text at the same time if it's big. You may have to regenerate same text over and over again so it would give you the best pronunciation. Please note that you only have 10k tokens a month. Use it wisely. Or consider to subscribe. Or use multiple Google accounts.🙈

4. When you are happy with the message, you go to RVC and make the character of your choice to say it with your voice model.

5. Then you just put all the parts of the message together with something like Audacity or other audio/video redactor of your choice.

And that's pretty much all. You can find a lot of guides for all the tools I mentioned on YT. Just search for the resent ones, because everything changes very quickly.

Hope it was at least a bit helpful?? If you have any questions please ask away. If you can't find a good guide, let me know, I'll find the best one for ya.

P.S. If you want to create AI COVERS, you also need Ultimate Vocal Remover (please find a guide on YT or ask me, I'll find one for ya).

Or you can use just this only tool Replay (But you won't be able to fix or edit the cover. I use it just for test usually.)