FBXL Social

Anyone aware of any good deepfake tts? Not for anything nefarious, I'd like to make an audio book version of my book and figure it'd be easier to get the computer to do the reading than to sit and hammer through it myself.

I found one called descript. It does two things, first it will split up a recording of speech into the individual words and allow you to edit as if you're just editing a text file. The other thing that includes is a deep fake text to speech. Now for ethical reasons, it forces anyone who's speech is being synthesized to say a whole thing that they're okay with their speech being synthesized, but the results were shockingly good. I uploaded a bunch of myself reading and a bunch of myself narrating the videos that I've done for work, and it was able to train the model using that, and I had to tell the people listening to the resulting recording that it wasn't me, because you really couldn't tell.

I found one open source deepfake thing, but there was absolutely no way I was going to take all the steps it was going to in order to make it work.

I'm always disappointed when I have to rely on closed software.