Elevenlabs.io - AI Generate anyone's voice with a minute of audio saying how much they hate trannies

[anustart76]

Has anyone been playing with the mrq tortoise repo lately? It seems like the training's gotten significantly worse lately.

 

[Keeper of Locks & Keys]

Has anyone been playing with the mrq tortoise repo lately? It seems like the training's gotten significantly worse lately.

That's what I've been trying to get to work. But it's not worth the hassle atm imo, it breaks all the time.

I'm waiting/looking for a way to have it or another solution run as good as Stable Diffusion, locally.

I can't imagine how this will turn given how litigious the Japanese are. Just look at the Abridged phenomenon. Never a clearer instance of fair use and boom! DMCA'd to hell.

DMCA takedown will never delete the source and stuff can always be re-uploaded or distributed in another way. The only question will be on how to reach your intended audience.

 

[Muttley]

 

[Puff]

Has anyone been playing with the mrq tortoise repo lately? It seems like the training's gotten significantly worse lately.

My install just died on update (no GPU detected, says to reinstall pytorch. Same issue on a clean install). Not sure if I've done something stupid or if there's a version compatibility issue in the repo. There are weird version errors on install that make me think there might be.

 

[anustart76]

That's what I've been trying to get to work. But it's not worth the hassle atm imo, it breaks all the time.

I'm waiting/looking for a way to have it or another solution run as good as Stable Diffusion, locally.

I think this is about as close as we'll have for a while. I think I've got it working, the real problem I had is that the usage info isn't really great on telling you what to do based on the amount of sample data you have.

My install just died on update (no GPU detected, says to reinstall pytorch. Same issue on a clean install). Not sure if I've done something stupid or if there's a version compatibility issue in the repo. There are weird version errors on install that make me think there might be.

Seems to be working on my machine. You might just have to uninstall pytorch and reinstall, that seems to be an issue. Make sure you're in your venv when you do that though.

 

[Puff]

Once again, I have no fucking clue if any fix is working because motherfucking whisper from github will not download at a reasonable stability or pace. 10kbps. I'm downloading it separately so I can actually monitor progress... had one get to 2.8 MB and pretended to be fully downloaded. Current attempt is at 5.1. Unexpected EOF errors abound.

To install from the zip, do I just throw it in the appropriate folder or do I have to do something?

Edit: Ok, full reinstall and still doesn't recognize my graphics card. This is gay.

 

[anustart76]

Once again, I have no fucking clue if any fix is working because motherfucking whisper from github will not download at a reasonable stability or pace. 10kbps. I'm downloading it separately so I can actually monitor progress... had one get to 2.8 MB and pretended to be fully downloaded. Current attempt is at 5.1. Unexpected EOF errors abound.

To install from the zip, do I just throw it in the appropriate folder or do I have to do something?

Edit: Ok, full reinstall and still doesn't recognize my graphics card. This is gay.
View attachment 4930534

You do have an Nvidia card, right?

 

[Cowboy Kim]

You do have an Nvidia card, right?

Very well could, this same shit happened to me with stablediffusion (1050ti on Windows 7.)

 

[IamnottheNSA]

Might as well post this here too.

 

[Puff]

You do have an Nvidia card, right?

Yes, and I've ran it before on gpu before a recent update. My initial setup problems were all failed downloads that the included installer wasn't handling properly, but not sure what this one is

 

[IamnottheNSA]

By the way, if you want to try out 11.ai, but don't want to give them your card info, there are various payment systems that offer "virtual bank card" services that don't require verification, like OlduBil. All you'll need is a burner phone number and an email.
Just make sure you don't go past the limit (for OlduBil it's 2750 Lira (~$140) a month - more than enough for an 11.ai subscription), because if you go over it they'll block your account and you'll have to verify it.

 

[IamnottheNSA]

I was always bummed that Boyd "Doc" Britton only ever voiced four Francis E. Dec rants, he absolutely nailed them.
Well, it's no longer a problem now.

 

[MrJokerRager]

 

[Sneeds]

Nulls big announcement:

You predicted the future

 

[Puff]

Yes, and I've ran it before on gpu before a recent update. My initial setup problems were all failed downloads that the included installer wasn't handling properly, but not sure what this one is

Reinstalled a third time and it worked. I don't fucking know. I don't fucking know. I don't fucking know.
I did OOM by not refreshing the voice list and telling it to rebuild latents. Tried to allocate 300 GB of VRAM.

 

[Sneeds]

Queen Kefaaalsss Is Freeeeeeeee

View attachment 4948365

Keffals reading the trans vengeance post

 

[Chimplord1997]

Is there a self-hosted alternative to elevenlabs that comes close to its quality? I want to make personal audiobooks or read articles to myself but there's virtually no good onboard TTS software for sale anymore, everything is cloud based and has stupidly low character limits. Even elevenlabs is 300+ a month for 40-ish hours of reading.

 

[Puff]

Is there a self-hosted alternative to elevenlabs that comes close to its quality? I want to make personal audiobooks or read articles to myself but there's virtually no good onboard TTS software for sale anymore, everything is cloud based and has stupidly low character limits. Even elevenlabs is 300+ a month for 40-ish hours of reading.

Tortoise-based AI Voice Cloning can be good if you've got a very clean dataset, but it has to be clean and you need a longer set than 11labs.
I did the Unabomber manifesto earlier in the thread as Peterson if you need an example.

 

[TheGuntinator]

Well, our celebrities will live forever as AI ghosts now, which is something. The first ones to do it will be the ones remembered for it. Look forward to having the A-list actors from the last twenty and next twenty years last literally forever.

Dio is already a Las Vegas Hologram.

Can someone who paid for this replicate Barik's voice? IDK what I want him to say, I just want to know it can be done. https://youtu.be/2XhbjQl2EMU

 

[Puff]

Dio is already a Las Vegas Hologram.

Can someone who paid for this replicate Barik's voice? IDK what I want him to say, I just want to know it can be done. https://youtu.be/2XhbjQl2EMU

Yeah, I could do some work and make it better, but here's the couple of click try. I just cut out the woman talking and a few of the shouts that I noticed.

Edit: and this was done locally. Probably better results on a paid service.