OpenAI claims its software can clone your voice from 15 seconds of you talking
- Reference: 1711990831
- News link: https://www.theregister.co.uk/2024/04/01/openai_voice_clone/
- Source link:
Described as being a "small model" that uses a 15-second clip and a text prompt to generate natural-sounding speech resembling the original vocalist, OpenAI said it's already been testing the system with several "trusted partners." It has provided purported samples of Voice Engine's capabilities in [1]marketing bumf emitted at the end of last month.
According to OpenAI, Voice Engine can be used to do things like provide reading assistance, translate content, support non-verbal people, help medical patients who've lost their voices regain the ability to speak in their own voice and expand access to services in remote settings. All those use cases are demoed and have been part of the work OpenAI has been doing with early partners.
[2]
News of the existence of Voice Engine, which OpenAI said was developed in late 2022 to serve as the tech behind ChatGPT Voice, Read Aloud, and its text-to-speech API, comes as concerns over voice cloning have reached a fever pitch of late.
[3]
[4]
One of the most headline-grabbing voice cloning stories of the year came from the New Hampshire presidential primary in the US, during which [5]AI-generated robocalls of President Biden went out urging voters not to participate in the day's voting.
Since then the FCC has formally [6]declared AI-generated robocalls to be illegal, and the FTC has issued a [7]$25,000 bounty to solicit ideas on how to combat the growing threat of AI voice cloning.
[8]
Most recently, former US Secretary of State, senator and First Lady Hillary Clinton has warned that the 2024 election cycle will be " [9]ground zero " for AI-driven election manipulation. So why come forward with another potentially trust-shattering technology in the midst of such a debate?
[10]OpenAI tweaks its fine print, removes explicit ban on 'military and warfare' use
[11]Microsoft seeks patent for tech to put words into your mouth
[12]Daughter of George Carlin horrified someone cloned her dad with AI for hour special
[13]You may have heard about AI defeating voice authentication. This research kinda proves it
"We hope to start a dialogue on the responsible deployment of synthetic voices, and how society can adapt to these new capabilities," OpenAI said.
"Based on these conversations and the results of these small scale tests, we will make a more informed decision about whether and how to deploy this technology at scale," the lab added. "We hope this preview of Voice Engine both underscores its potential and also motivates the need to bolster societal resilience against the challenges brought by ever more convincing generative models."
To assist in preventing voice-based fraud, OpenAI said it is encouraging others to phase out voice-based authentication, explore what can be done to protect individuals against such capabilities, and accelerate tech to track the origin of audiovisual content "so it's always clear when you're interacting with a real person or with an AI."
That said, OpenAI also seems to accept that, even if it doesn't end up deploying Voice Engine, someone else will likely create and release a similar product - and it might not be someone as trustworthy as them, you know.
[14]
"It's important that people around the world understand where this technology is headed, whether we ultimately deploy it widely ourselves or not," OpenAI said.
So consider this an oh-so friendly warning that, even if OpenAI isn't the reason, you can't trust everything you hear on the internet nowadays. ®
Get our [15]Tech Resources
[1] https://openai.com/blog/navigating-the-challenges-and-opportunities-of-synthetic-voices
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZgsufI7XgEH5Szefn-GGbQAAAIs&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZgsufI7XgEH5Szefn-GGbQAAAIs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZgsufI7XgEH5Szefn-GGbQAAAIs&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.theregister.com/2024/01/23/robocaller_biden_new_hampshire/
[6] https://www.theregister.com/2024/02/08/sorry_scammers_the_fcc_says/
[7] https://www.theregister.com/2024/01/05/ftc_voice_cloning_solution/
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZgsufI7XgEH5Szefn-GGbQAAAIs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://www.theregister.com/2024/03/29/hillary_clinton_election_ai/
[10] https://www.theregister.com/2024/01/15/ai_in_brief/
[11] https://www.theregister.com/2024/02/08/microsoft_patent_speach/
[12] https://www.theregister.com/2024/01/12/george_carlin_comedian_cloned/
[13] https://www.theregister.com/2023/06/28/adversarial_ai_audio_authentication_research/
[14] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZgsufI7XgEH5Szefn-GGbQAAAIs&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[15] https://whitepapers.theregister.com/
Re: 15 seconds
Check out the original article on this at The Verge. It has 3 synthesized voice samples generated from 15 secs input...really, really scary.
Re: 15 seconds
"3 shampla gutha sintéiseithe ginte ó ionchur 15 secs...dáiríre, scanrúil" ... I'm confident that Brendan Behan (icon) would have said "I am an AI atheist." (in a Youth Custody Centre these days) ... So these days Brendan would have had to be writing a book titled "The Youth Custody Centre Boy" but an AI voice would never have made it sound as unique a "Borstal Boy"
Re: 15 seconds
There are other AI models out there that do similar things. OpenAI definitely isn't the first.
Hollywood 2
I can imagine that within 2 years at least 90% of all videos and recording will be AI generated.
In other words it will become the 2nd Hollywood, full of deception, false dreams and lies.
Re: Hollywood 2
> the 2nd Hollywood, full of deception, false dreams and lies.
(tongue -> cheek) So not like the current Hollywood at all then?
April 1
I am really hoping the date of this article is significant but I fear not.
Re: April 1
Was posted in the afternoon, so if it was it was late :0
Re: April 1
There are other AI models out there with similar claims which do work to varying degrees, so I would rate it on the same level of credibility as Google announcing that they've launched a spreadsheet program.
Where it likely will go wrong is if for example an English person provides a voice recording and asks it to convert it to Donald Trump's voice, it could sound like Donald Trump speaking in an English accent.
Also if there is a word that a particular person always says in a weird way, like for example Nigella Lawson's meekrowavey, it is probably not going to replicate that, even if it is in the sample.
15 seconds
"Described as being a "small model" that uses a 15-second clip and a text prompt to generate natural-sounding speech resembling the original vocalist"
I just wonder how close they can generate a specific accent, like Geordie, Scouse etc. from a mere 15 seconds of speech. Although I guess that using Siri or Alexa they'll have hours of eavesdropped speech to use. We should really start worrying when it can deduce how you pronounce "scone" without having heard you ordering tea in Fortnums.