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

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
Since John Goodman brought up Waits v. Frito-Lay, I was looking around for similar cases, and saw Bette Midler v Ford. Similar case to Waits: Midler was approached to do a song for Ford. She refused. Ford bought rights to the song and hired a soundalike (one of her backup singers, in fact), to sing. Apparently the imitation was real good.

She sued, and lost at summary judgment at the trial court. Ninth Circus picked it up. They found since Midler's voice (she was a well known singer) was an important part of her celebrity status, misappropriating it was tortious. They did make a note that the voice had to be an important facet to the plaintiff's identity, and that her voice was widely known.

This could be a set up for some really weird legal precedents.

For example, while sure, Lindbeck uses her voice for her work, she doesn't use one voice. Since she fits her voice to different roles, it could be argued that her voice isn't a distinct identity.

It also has to be widely known. This was brought up back in the Vic Mignogna case, whether an anime VA was widely known enough to be considered a public figure. It was never conclusively settled since Ty Beard is a moron, but there could be arguments for an against any given voice actor (specifically, a voice actor's voice) being considered widely known enough to be protected.

This of course assumes there is tangible damages, which a parody song doesn't give.
 
For example, while sure, Lindbeck uses her voice for her work, she doesn't use one voice. Since she fits her voice to different roles, it could be argued that her voice isn't a distinct identity.
And this in fact is the primary stumbling block (at least for a suit in California under their right of publicity laws that for obvious reasons favor show biz).
 
I recently setup RVC, its annoying to find models for it because discord niggers set it all up on a discord instead of hosting them on literally any other site. It's pretty good and is surprisingly fast to generate audio and even train models when using a GPU, and I haven't even run into VRAM issues like with Stable-Diffusion.

Here is my favorite one so far, had to manually add in the radio beeps:


I haven't done very much with it yet since I'm not sure what to even do, so I've just been testing various models with the ambatukam audio and the thug shaker audio, and some Dagoth Ur voicelines since I had all his lines set up in a dataset already. It works really well with clean voice acting from a game, you could probably use it to re-voice entire characters, as long as the lines are delivered cleanly.




I even trained it on the 'Welcome to the Cumzone' guy, since the guy provided the source audio and its already cleaned and in a single file, so I just truncated silence and then threw it in and it took about 45 minutes to train 300 Epochs on a 3070ti

I might make a Quake Live voice pack with this actually, its a good announcer voice.

Here are some examples on clean game voice acting, Dagoth Ur lines because I had them available




I've noticed that it does pick up the accent of the original voice, so with the thug shaker audio it becomes southernish, it also just sounds messy because the audio is not very clean. I only did a basic noise removal on it.


I think the combine one sounds so good is because it hides the AI jank behind the compressed and noisy voice filter. Also Dagoth Ur doesn't sound the best in this model because there is only ~2 minutes of audio, the Cumzone guy has ~16 minutes of audio to use. Don't have any Dagoth examples because they all sounded kinda bad. I haven't tried singing yet, the webui has a tool for vocal isolation and vocal removal, so it should be easy to do.
 
RVC really is pretty cool, I was able to train a pretty decent model in about 90 minutes.

I trained a model to reproduce the singing voice of Kevin Conroy/Batman (RIP). I always wanted to hear him sing Snake Eater, since he had a perfect theatrical singer type of voice that I always thought would suit a male version of the song perfectly. Sadly, he's gone so we'll never hear him sing it for real, but the AI came up with a surprisingly good result. The biggest weakness seems to be that it copies the original singer's accent almost exactly, or at least tries to, so the accent output doesn't match the new voice's source. It also doesn't like long silences and inserts weird noises, but those can be edited out easily.

I had to use the newer rmvpe option instead of harvest or crepe, because those kept causing some kind of artifacting that sounded like a voice crack.

Here's the solo version. I had to clean up the audio a little and add some effects to get it to sound the same way the original does.


I also made a duet with Cynthia Harrell's original singing, which came out beautifully.


For comparison, here's an actual track that Conroy recorded a few years ago of him singing "Am I Blue"


And since I know you Jokers (pardon the pun) will ask for it, here's the output with the voice crack effect:
 
Neco Arc makes her break into jazz sinatra.png


instrumental:

vocals:
 
Do AI covers fall under copyright strikes as this one by Frank Sinatra singing some Lana Del Ray song did.


And I should have downloaded it last night as it got yeeted today.
 
Do AI covers fall under copyright strikes as this one by Frank Sinatra singing some Lana Del Ray song did.


And I should have downloaded it last night as it got yeeted today.
for now it's still a grey zone, but afaik universal are among the biggest niggers about it (and among the most strike-happy) .
probably also the first and most prolific who milk all the "content" they own once a legal foundation is established. see:

there was also a lawsuit where apparenly similar chords are enough these days (not yet at least):

one of the angles they try to come after it are: "it was trained on copyrighted material", which means courts have to decide how transformative it really is to possibly fall under free use, and "muh misinformation" (no idea what that has to do with music, but retards gonna retard), which is a much juicer aspect they really like because it gives them more control.


something else youtube recommended me after watching some of the embedded videos
 
for now it's still a grey zone, but afaik universal are among the biggest niggers about it (and among the most strike-happy) .
probably also the first and most prolific who milk all the "content" they own once a legal foundation is established. see:

there was also a lawsuit where apparenly similar chords are enough these days (not yet at least):

one of the angles they try to come after it are: "it was trained on copyrighted material", which means courts have to decide how transformative it really is to possibly fall under free use, and "muh misinformation" (no idea what that has to do with music, but retards gonna retard), which is a much juicer aspect they really like because it gives them more control.


something else youtube recommended me after watching some of the embedded videos

Seems like he has reuploaded it now.

 
  • Like
Reactions: ZMOT
Lately its been a pain in the ass for me to do RVC. I know how to seperate vocals and everything. My problem is figuring out how to get the AI voices they have on discord downloaded into it

I know how to run the browser RVC on google docs or whatever, The ai models I get from Discord just dont seem to work because either I link it directly from discord and it doesn't work, or I download the whole thing and it doesn't work.

And as for the RVC I actually download into the computer, its been a pain in the ass to actually start the dam thing. Nobody has good instructions on it.

There's just to many dam conflicting youtube guides that either leave out the one tiny ass thing I need to know, or they spend 10 mins on how to create an album when I already have something someone else made and dont want to spend an hour on it.
 
  • Like
Reactions: Toolbox
Are there any updates for TortoiseTTS? Whether or not it's gotten up to speed to Eleven Labs yet? I kinda wanna use it to clone a character's voice since they only say one line and that's it, and I figured if I use something like Tortoise for cloning I could use that for a dataset to then train on RVC and make it say whatever I want in a more natural and controllable way.
Lately its been a pain in the ass for me to do RVC. I know how to seperate vocals and everything. My problem is figuring out how to get the AI voices they have on discord downloaded into it

I know how to run the browser RVC on google docs or whatever, The ai models I get from Discord just dont seem to work because either I link it directly from discord and it doesn't work, or I download the whole thing and it doesn't work.

And as for the RVC I actually download into the computer, its been a pain in the ass to actually start the dam thing. Nobody has good instructions on it.

There's just to many dam conflicting youtube guides that either leave out the one tiny ass thing I need to know, or they spend 10 mins on how to create an album when I already have something someone else made and dont want to spend an hour on it.
Don't trust any videos you see on making AI shit with RVC, chances are it's already outdated. This doc here is pretty much always kept up to date on using the collab: https://docs.google.com/document/d/13_l1bd1Osgz7qlAZn-zhklCbHpVRk6bYOuAuB78qmsE/edit?usp=sharing
As well, this is the current link used by many to make ai covers: https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing
As for problems with downloading, you made sure to copy the link as is and pasting it in this field right?
iMjv3UtQfC.png
I know for a fact that this bitch is super fucking case sensitive and you might've added a space at the end that ended up fucking with the download. Either that, or you ran out of GPU usage for the day and switched to using a CPU instead, which by that point you'd be better off switching to a different account to use a new GPU or wait until it gets renewed for tomorrow instead.
 
Have they cracked down on this yet? Elevenlabs I mean.
I don't think so but for anyone with some skill I think elevenlabs is outdated anyway. Open source options have beat it out in several ways. I think it's 'make your own voice' thing hasn't been directly replicated by these but that was a shitty gimmick anyway.
 
I don't think so but for anyone with some skill I think elevenlabs is outdated anyway. Open source options have beat it out in several ways. I think it's 'make your own voice' thing hasn't been directly replicated by these but that was a shitty gimmick anyway.
Huh. From my limited perspective it feels like they're trying to grab as much money as they can before getting the boot on their neck, sort of like Napster or even Facebaook, but unlike napster they are smart enough to self-regulate and update their TOS before that happens before something like an AI Dungeon scandal breaks out.
 
Back