OpenAI can clone voices based on just 15 seconds of speech

RoiBox

Elite Member
Executive VIP
Jr. VIP
Joined
Apr 4, 2009
Messages
12,550
Reaction score
62,539
SOME OF the visions of the future in the sci-fi series Black Mirror are already technically possible, more or less. In the 2013 episode "Be Right Back" starring Hayley Atwell and Domhnall Gleeson, a grieving girlfriend digitally recreates her boyfriend after he died in a car accident.

By scanning all the boyfriend's text messages, sent emails and posts on social networks, his personal way of expressing himself can be imitated by a chatbot. The boyfriend's voice is cloned from existing audio recordings, like videos. The technology required for this is available today through generative AI and language models.

https://openai.com/blog/navigating-the-challenges-and-opportunities-of-synthetic-voices
Cool.
 
Yeah, think; wife calling to check up on you "as" another cute girl.
And I'll answer in my mother-in-law's voice, telling her how lucky she is to have me, and to leave me alone, cause I've worked until just now; And that it should be fair to prep me some cold beers when I'll get home (bonus.).
 
XTTS can do this right now based of a 10 second clip or if you are hardcore and got hours worth of audio you can clone someone's voice perfectly with Tortoise TTS (which is the same thing i think)

Unfortunately it pronounces everything perfectly instead of prolonging certain letters and it lacks emotion to anything it says.
 
To me that is scary, amazing and scary to a dark world were every time is difficult to trust everything you listen or see through any electronic device
 
To me that is scary, amazing and scary to a dark world were every time is difficult to trust everything you listen or see through any electronic device
The world of AI is growing more frightening. I can't conceive how quickly cybercrime will increase. Out of 100%, 20% would utilize this to deflect people.
 
voice cloning is cheap af. almost every other company is launching a voice cloning service (with real emotions) even. though openai has kept its service in private their are plenty of very cheap options in the market
 
"OpenAI announced this week that it has developed a tool that clones a person's voice simply but will not release it to the general public as a sign of its "cautious and informed" approach."

They did stopped it and is acting ResPOnsIbly.
 
It actually doesnt matter even if openai decides to keep this in private indefinetly. there are very good open source models available and its very cheap even if you decide to pay for it.
 
Back
Top