I'm going to assume somebody's done this, so...who's tried this? Sites like C.AI let you create a voice off of a 15 second clip.

Has anybody taken, say, a video from a real model and used it to create a voice?

Any tips for doing this?

Recommendations for sites for ripping from Youtube and/or isolating speech from videos with backing music?

How much trouble can you get in for doing this?

real models usually have the most cringy voices, better to use professional VA or smut recordings... unless there's a specific model whose voice gets you going...

I've never done AI audio before. Are there any services which can do stuff like moaning and grunts?

>>26869

good question. c.ai is particularly bad at even just interjections or stammering, so...

>>26773 (OP)
To get great audio samples to clone a character's voice, you must check for background noises, always check that there are no sound effects nor music playing in the background while the character talks.

Here's an audio sample that I used to clone Sara Bellum's voice. Notice that there's nothing playing in the background? It's only Ms. Bellum's voice.

And here's the result after using that file in ElevenLabs.

>>26773 (OP)
https://youtu.be/2-ilbFBeZk4?t=14

Here's the clip where I got Ms. Bellum's voice.

>>26915

awesome work. have you had any success using AI isolation tools to clean up a noisy file before cloning?

>>26773 (OP)
I use tortoise still, but I hear xtts2 is much better. you don't need much, just a few seconds to be able to get the latents, but quality does matter a lot.
use yt-dlp/ffmpeg to rip the audio, then cleanup in your DAW of choice.
no trouble if you don't share. I have a pretty extensive library of voices, and I can't imagine wanting to make them available anywhere else, not that there'd be any interest.

>>27055
Ai tools? I use Adobe Audition to clean and enhace the audio clip before uploading it to Ai.