Changing voice of one actor in an audio drama, with auto detection of that voice, not manually.

[lime patio]

As in title:
I bought a 2 hour long audio drama (fully casted) in mp3 and the main character is voiced by an impersonator rather than a real actor that played the same character in tv series. And he sounds awful. I want to change his voice using an ai to a real voice actor. But heres the catch. I dont want to spoil myself a plot. So i need a method or program to automaticly find a every instance of voice of that one actor in a whole file (maybe based on a short voice sample idk), and replace only that voice with AI. Is this possible?

[prime rain]

What's your PC GPU

You can just give RVC the audio file, then the trained model of the original voice actor, without needing to check the file manually

However, you need a trainer model of the original voice actor

[lime patio]

GTX 1060, very old graphic card

But come on, that's only 2 hours of audio, selectively around 20 minutes maybe, it can't take that much time

[lime patio]

Original you mean the bad impersonator from audiobook or the real actor I want to hear?

[stable narwhal]

the opposite

[prime rain]

You would need a model of the original voice actor you want to hear

[lime patio]

Okay, I used command /find and I already found it, sounds good enough

[prime rain]

Eh, it could run it locally, it should work just might be slow

If you want, there's cloud (remote good PC), which would be faster with a better GPU, though it's tied to a limited time

[lime patio]

Don't care, I don't need it right away, how long could it take?

[prime rain]

Do you want it to run on your PC (locally), or use a cloud service?

[prime rain]

Sorry but we can't really estimate that

[lime patio]

Even with a large margin, just indicatively, how long? Hour, hours, days?

[lime patio]

I would rather choose my pc

[prime rain]

Maybe an hour or more, but I'm not sure, I can tell you it won't take a day+ though

[prime rain]

Alright

As you got a good PC, you can use RVC locally, you can choose between:

• Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
• Mainline: The original RVC

Personally, I would suggest Applio

[lime patio]

Okay, thank you, I will do it tomorrow or at the weekend, thanks for help. I will be updating this thread in case of needing help. Once again, thanks.

[prime rain]

Alright yw

[lime patio]

what exactly do you mean voice model? There is a good model on Weights but i dont know how to download it

[prime rain]

To Download a Model from Weights.com:

1. Login
2. Click the 3 dots at the right of the image of the model
3. Click download
4. Download Anyways
5. Unzip the zip, and you might wanna rename the pth and index since all models on weights are renamed as 'model'

[cursive spear]

If you have an audio file with two different persons speaking

I dont know any model that would separate them into two different files and preservice the timings

and I think that's what you want

all the advice above would replace ALL voices in the audio with the same voice

and I dont think that's what you want

[lime patio]

Well, that's exactly what I want to do XD

[stable narwhal]

there are no good models for separating ppl speaking together, mostly compromising or ruining the result quality, or even not effective

[prime rain]

@lime patio btw how's it going? You could also try some karaoke model to separate the vocals

[lime patio]

i just listened to the audio drama normally, after some time i managed to get used to the impersonator voice and the plot was great so it wasnt that much of an issue

altrought i still experimented in Weights with Capaldi AI

i might make another post later to ask about some technical stuff with ai , but this little project i am making is far from being even started, i have a lot of things going around me now

[prime rain]

so should this post being kept open ?

[lime patio]

oh, sorry i didn't knew i should close it. of course it can be closed.