>New AI technology drops (eg. Voice deepfakes)
>Everyone freaks out, it gets media attention
>People are convinced this is the end of voice actors and in 6 months the tech will be settled
And then for like 5 years nothing happens. Open source solutions get slightly better but it never really takes off
And then suddenly
>2 polish guys make some random saas web app
>It blows literally everything out of the water
>Now people give a shit about voice deepfakes again
Why wasn't there a clear predictable progression to now, why did it seem to fade away despite new open source stuff coming out all the time? Is it really just that it finally cracked the nut of genuine sounding expressions/vocal tones?
Don’t worry we still have a long way to go
It will get better if we are autistic enough to give it goodwork
real?
Look at that cat bird. This cannot last. Our world is collapsing.
Science isn't stable progress at all, it's sudden leaps because you don't get it until you do, after which it just works.
it was made for mod for a 23 year old game that they didn't want to pay voice actors for so they created a voice synthesizer
dangerously based
what game?
gothic
There has been incremental progress to now, but elevenlabs represents one of the first turnkey solutions any idiot can use which explains the explosion in interest.
the new voice AI is better than anything before it and you can actually use it unlike the shit before.
>the new voice AI is better than anything before it and you can actually use it unlike the shit before.
You cant "actually use it" because its not local and they are monitoring everything you write in to it
you can it just like chatgpt or stable diffusion for lots of people it sucks that you can't run it locally but there will be a opensource alternative in due time.
This. Look into the devs connections to the feds. It's a honeypot at worst, a data harvester at best.
Is there anything useful you can do with this information?
export const firebaseConfig = {
apiKey: "AIzaSyBSsRE_1Os04-bxpd5JTLIniy3UK4OqKys",
authDomain: "xi-labs.firebaseapp.com",
databaseURL: "https://xi-labs-default-rtdb.firebaseio.com",
projectId: "xi-labs",
storageBucket: "xi-labs.appspot.com",
messagingSenderId: "265222077342",
appId: "1:265222077342:web:3acce90d1596672570348f",
measurementId: "G-82RG1PXYVW",
};
export const serverUrl = "https://api.elevenlabs.io";
export const redirectUrl = "https://beta.elevenlabs.io";
export const stripePublishableKey =
"pk_live_51M07hSLmdOdiMXBscAs5C18VwtmKFR2bBXIpYn244iQ9taunLhcGIgB6cZ4m4X5Hr5tpZu265LQ0mGRDYhQuGVd100bdpNnycY";
export const envType: "local" | "dev" | "prod" = "prod";
Lmfao is that public on their website? URL or it's fake
you guys have obviosuly never used firebase
>is there anything you can do with public api keys?
OP you sound retarded
If you can't tell the difference between whatever it existed 5 years ago and the cutting edge deep learning shit that is being implemented today you are just cattle pol conspirationist
I’m sure governments have had this tech for at least a decade
You haven't been keeping track. gayMAN constantly publishes papers. But they never release anything exciting. Not even for demo use as a SaaS.
https://google.github.io/tacotron/
why isn't there an AI that can take translate japanese language and provide subtitles? This is something we should have by now.
Because that would be going backwards when we're on the verge of having AI that can translate and provide dubs in the original VA's voice.
anyone see their new payment plans? 500k tokens for $121
2 million tokens for $440
this shit must have a generation cost equivalent to 2014 dogecoins for there to be such a price discrepency and for the free tier to still be available
someone will figure it out in no time and make their own open source version
even if it went open source or got leaked it probalby requires 100GB of VRAM to load
the new status quo is to be ramcucked
128g RAM reporting in
VRAM
V
sexy glasses apu t. femanon
your going to start seeing $10k+ "hobbyist" builds for this shit with multiple 4090s and custom GPUs that connect to 240v outlets
we're going back to the 70s-80s
good photoshop exists since ages.
It's more a question of media reporting. News media is very saturated, and people hear constantly about a number of things that are going on in the world. Development on voice models was developing in the backround. Once a model was good enough it was launched and got some traction. For all we know there were even better models in the meantime that just didn't get the same online traction so we didn't hear about them. It's not like the development stopped until some random people solved some important problem, more that people didn't hear about it for a while and now they hear about it.
/ai/ board when?
does tacotron still work i remeber there bing a local install for that
Why wouldn't it work.