Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror
×
AI

OpenAI Didn't Copy Scarlett Johansson's Voice for ChatGPT, Records Show (msn.com) 74

The Atlantic argued this week that OpenAI "just gave away the entire game... The Johansson scandal is merely a reminder of AI's manifest-destiny philosophy: This is happening, whether you like it or not."

But the Washington Post reports that OpenAI "didn't copy Scarlett Johansson's voice for ChatGPT, records show." [W]hile many hear an eerie resemblance between [ChatGPT voice] "Sky" and Johansson's "Her" character, an actress was hired in June to create the Sky voice, months before Altman contacted Johansson, according to documents, recordings, casting directors and the actress's agent. The agent, who spoke on the condition of anonymity, citing the safety of her client, said the actress confirmed that neither Johansson nor the movie "Her" were ever mentioned by OpenAI. The actress's natural voice sounds identical to the AI-generated Sky voice, based on brief recordings of her initial voice test reviewed by The Post...

[Joanne Jang, who leads AI model behavior for OpenAI], said she "kept a tight tent" around the AI voices project, making Chief Technology Officer Mira Murati the sole decision-maker to preserve the artistic choices of the director and the casting office. Altman was on his world tour during much of the casting process and not intimately involved, she said.... To Jang, who spent countless hours listening to the actress and keeps in touch with the human actors behind the voices, Sky sounds nothing like Johansson, although the two share a breathiness and huskiness. In a statement from the Sky actress provided by her agent, she wrote that at times the backlash "feels personal being that it's just my natural voice and I've never been compared to her by the people who do know me closely."

More from Northeastern University's news service: "The voice of Sky is not Scarlett Johansson's, and it was never intended to resemble hers," Altman said in a statement. "We cast the voice actor behind Sky's voice before any outreach to Ms. Johansson. Out of respect for Ms. Johansson, we have paused using Sky's voice in our products. We are sorry to Ms. Johansson that we didn't communicate better..."

[Alexandra Roberts, a Northeastern University law and media professor] says she believes things will settle down and Johansson will probably not sue OpenAI since the company is no longer using the "Sky" voice. "If they stopped using it, and they promised her they're not going to use it, then she probably doesn't have a case," she says. "She probably doesn't have anything to sue on anymore, and since it was just a demo, and it wasn't a full release to the general public that offers the full range of services they plan to offer, it would be really hard for her to show any damages."

Maybe it's analgous to something Sam Altman said earlier this month on the All-In podcast. "Let's say we paid 10,000 musicians to create a bunch of music, just to make a great training set, where the music model could learn everything about song structure and what makes a good, catchy beat and everything else, and only trained on that... I was posing that as a thought experiment to musicians, and they were like, 'Well, I can't object to that on any principle basis at that point — and yet there's still something I don't like about it.'"

Altman added "Now, that's not a reason not to do it, um, necessarily, but..." and then talked about Apple's "Crush" ad and the importance of preserving human creativity. He concluded by saying that OpenAI has "currently made the decision not to do music, and partly because exactly these questions of where you draw the lines..."
This discussion has been archived. No new comments can be posted.

OpenAI Didn't Copy Scarlett Johansson's Voice for ChatGPT, Records Show

Comments Filter:
  • by LondoMollari ( 172563 ) on Saturday May 25, 2024 @10:40AM (#64498281) Homepage

    Sky was the only usable voice for ChatGPT. The others are hard to listen to and it has decreased my use a bit. OpenAI needs to put this to bed, assert themselves, and get the Sky voice online. No more of this overly sensitive stuff. It has been proven it wasn’t a copy, it’s property they made, now they should use it (because it was the best).

    • Seems strange to say the only usable voice OpenAI can use is one that sounds very similar to a world famous actress who happened to famously voice an AI.

      I agree with you that if it ain't her then it ain't her so just go for it but that's just a weird proposition, like what circumstances of the model led to that result?

      • by g01d4 ( 888748 )

        Seems strange to say the only usable voice OpenAI can use is one that sounds very similar to a world famous actress who happened to famously voice an AI.

        That was my thinking as well, only similar -> extra publicity (good, bad or neutral) may have been a strong factor.

        • by Rei ( 128717 )

          It's clear that Altman at the very least has no objection to soundalikes, and had recognized the similarity. Which is part of what makes his stronger stance on AI music so weird.

          He concluded by saying that OpenAI has "currently made the decision not to do music, and partly because exactly these questions of where you draw the lines...

          Thankfully nobody else will... [youtube.com] ;)

          • It's clear that Altman at the very least has no objection to soundalikes, and had recognized the similarity.

            Are people perhaps confusing an individual person's voice with the direction that voice actresses received? For example "deliver the lines in a friendly, bubbly style". Personally from listening to both voices side by side I think the similarity is in the performance style not the voice itself. They both sound like a young college aged women having a conversation that is on the verge of becoming flirty. Not there yet but seeming to head that way.(*)

            (*) For those of you with limited experience talking wit

        • Seems strange to say the only usable voice OpenAI can use is one that sounds very similar to a world famous actress who happened to famously voice an AI.

          That was my thinking as well, only similar -> extra publicity (good, bad or neutral) may have been a strong factor.

          Although, I can think of a more obvious actress to imitate if they were thinking of exploiting the Streisand effect [wikipedia.org] for their publicity ... :-)

      • > like what circumstances of the model led to that result?

        Well, how much variation is there in human voices really? Can you distinguish 100 people uniquely by their voice? 1000? If 1000 hollywood stars are granted the right to copyright their voices, do the rest of us go with sign language?

        • 1000? Probably not but 100+? absolutely. I bet most anyone could recognize at least 50 to 100 actors by voice alonecon top of their coworkers and family and friends. Isn't there some theory that our brains can only maintain real familiarity with like a 150 people?

          It's not really about copyright, I don't think you can copyright "something you are" only something you create right? Like I couldn't copyright my voice but I could copyright a character I created with a specific voice? I could be wrong but I

          • Isn't there some theory that our brains can only maintain real familiarity with like a 150 people?

            Recognizing people by voice is not quite the test we want here IMO. I recognize many of my coworkers because of their accents, turns of phrase, or the way one guy always extends the first sound in a sentence.

            For these voice assistants we're probably looking at clear pronunciation as the bare minimum, no accents, probably with further filtering on voices that "sound good". My intuition is that if you try to make a bunch of voice banks that vary in pitch/timbre, with the goal that you can present any two of t

      • Seems strange to say the only usable voice OpenAI can use is one that sounds very similar to a world famous actress who happened to famously voice an AI.

        Its not the voice, its the personality portrayed. A young, friendly, bubbly female. Its that personality that attracts attention.

        I agree with you that if it ain't her then it ain't her so just go for it but that's just a weird proposition, like what circumstances of the model led to that result?

        FWIW I heard side by side comparisons. It was obvious to me that it was two different women. However the personality portrayed by both voice actresses induced "flashbacks" of happy college days.

        • Its not the voice, its the personality portrayed

          Perhaps, but can someone copyright a personality?

          • by drnb ( 2434720 )

            Its not the voice, its the personality portrayed

            Perhaps, but can someone copyright a personality?

            That's the problem, and why OpenAI is likely in the clear. That personality is too common in the real world.

            Personally I expect this is all just Johansson making it crystal clear to the world she is not involved. So when the inevitable embarrassing or offensive things are said by the AI she will not be involved in the controversy. Its a preemptive PR move to distance herself.

      • by sjames ( 1099 )

        The circumstance was that the voice actress employed for the training data naturally sounds a lot like Johansson.

        If that is to be considered unacceptable, she is effectively enjoined from her profession.

    • This is exactly why I think the idea of "owning" a voice -- any voice, including your own -- is absurd. A recording of your voice, sure, but not the sound of your voice, accent, etc. I mentioned exactly this scenario here a few months back, specifically giving the example of a voice actor who sounds exactly like Morgan Freeman. Morgan Freeman would have no basis to claim ownership of it. Neither does this actress.

      A better analogy to this is like the guy who invented FM synthesis owning copyright on literall

      • Re: (Score:3, Informative)

        There's nothing absurd about "owning" a voice. It's just how business works. Johanssen spent many years making her voice stand out, it's part of her brand, and it has value. She has to defend it against impersonation and slander.

        Generally businesses sue other businesses who operate copycat services or products. For example, everyone knows how to make a burger, but McDonald's sues any unlicensed person who makes burgers that pretend to be big macs.

        • There's nothing absurd about "owning" a voice. It's just how business works. Johanssen spent many years making her voice stand out, it's part of her brand, and it has value. She has to defend it against impersonation and slander.

          And by this same logic: Does not the voice actress who actually modeled for this system ALSO have a right to sell her voice work? Even if she sounds similar to Scarlett Johansson? Should she be prohibited from earning a living as a voice actress because she sounds too similar to a more famous actress?

          • The actress totally has the right to work. OpenAI's strategy of choosing an actress who sounds like Johansson because it will remind users of the connection with a popular movie and make them think ChatGPT is as smart and capable as that movie portrays is exploitative. The inevitable racist and criminal babble that some users will tease out of ChatGPT in the form of sentences spoken by Johansson sound-alike will cause her reputational damage or worse. This is so obvious it should not need to be pointed ou
        • by mysidia ( 191772 )

          There's nothing absurd about "owning" a voice. It's just how business works
          Intellectual property is to protect Creations NOT the hours or years of work you put in.

          Put this way.. No matter how much businesses would Love to own an absolute exclusive right to a Voice: No such right exists, AND no such right would be just nor should ever be allowed to be created.

          We put an Amendment into the constitution to Allow exclusivity of SOME works of the mind such as books or songs you wrote, But Not abstract "ideas

          • The problem isn't the voice. The problem is OpenAI are cynically choosing to make their AI sound like Johanssen for marketing reasons. It's a stupid business strategy invented by naive techbros which opens them up to lawsuits in the future related to impersonation and reputation damage. It has nothing to do with any amendments in any land on earth.
        • by thomst ( 1640045 )

          martin-boundary pointed out:

          There's nothing absurd about "owning" a voice. It's just how business works. Johanssen spent many years making her voice stand out, it's part of her brand, and it has value. She has to defend it against impersonation and slander.

          Generally businesses sue other businesses who operate copycat services or products. For example, everyone knows how to make a burger, but McDonald's sues any unlicensed person who makes burgers that pretend to be big macs.

          Once again the seemingly endless ability of /. commenters to conflate copyrights with TRADEMARKS is what muddies the waters of this discussion. A work - a concrete, self-contained, unique product - is subject to copyright protection. A representation or symbol of a product, on the other hand is subject to protection under trademark law. You can't, for instance, copyright a name, but you CAN trademark it. (Thus: McDonald's.)

          Legally speaking, the major distinction is that copyright

        • There's nothing absurd about "owning" a voice. It's just how business works. Johanssen spent many years making her voice stand out, it's part of her brand, and it has value. She has to defend it against impersonation and slander.

          No, that's not how business works. As a matter of law, you can't patent or copyright naturally occurring things or properties. Hasbro was the first company to manufacture action figures, in the form of GI Joe. They couldn't secure a copyright on it though, because it was a representation of the human body. Instead they secured a copyright on what were originally manufacturing defects that now deliberately manufacture into the toys.

          You can own a recording of your voice and you can own a choreography of a dan

          • The discussion was about "owning" in quotes. The issue is that the OpenAI techbros think that they are safe to impersonate Hollywood A-listers with their ChatGPTs for profit, and your statements, while true, miss the point. OpenAI is doing something unsafe, because their AI is pretty dumb and easy to jailbreak. People will make it say all sorts of things that will damage Johanssen's reputation or worse if unchecked. She can't be okay with that and OpenAI will be sued for deliberately causing trouble.

            The

        • > For example, everyone knows how to make a
          > burger, but McDonald's sues any unlicensed person
          > who makes burgers that pretend to be big Macs.

          And if you made a cheeseburger with that extra patty and bun, then added thousand-island dressing diluted with mayonnaise, and sold it at your own burger stand; those lawsuits would get thrown and laughed out of court every time... so long as you don't sell them AS "Big Macs."

          • As we are already finding out in other areas of impersonation by machine, ordinary people don't tend to make a distinction between a real person and a deepfake. So if your cheeseburger is a "Big Nac" and has a logo with lemon green arches on a pink background, it's too close for an ordinary person of the public to distinguish from a real "Big Mac", so you will get sued even if you didn't claim it is an actual "Big Mac". OpenAI's deliberately close representation of Johanssen's voice falls into that category
      • What about your name?

        Recent news in Wa: "Democratic candidate Bob Ferguson pursued various hardball political and legal tactics to get two opponents who share his name out of the governorâ(TM)s race â" including an option the secretary of state said Washington election law doesnâ(TM)t allow."

    • Oh just not have a voice at all.

    • No more of this overly sensitive stuff? Sam will probably lose. Bett Midler was approached for a Superbowl ad years ago to use her voice and she refused. They hired a soundalike and she sued and won.

      Sam reached out to scarj 6 months and she declined. Then 48 hrs before release and without her responding they release it and he tweets one word, 'her'. The courts look at intent. Was it coincidence they sound similar or was their intent? The multiple offers and tweet show intent and Sam will lose.

    • by phayes ( 202222 )

      > It has been proven it wasn’t a copy

      No, it has most certainly not been proven. It has been asserted but that is not proof.

      OpenAI has been proven to have deemed that anything that they can use anything they can access on the Internet to train their AIs without any considerations. That there are no currently known tracks between Altman and the Sky voice resembling SJ's voice doesn't mean that he didn't pull Roberts aside and tell her "make it sound like SJ" or that there are communications that they

  • by Pinky's Brain ( 1158667 ) on Saturday May 25, 2024 @10:44AM (#64498289)

    He was so hands off he tried to personally do licensing negotiations. Altman should have let sleeping dogs lie.

    • This.

      Even if Altman recognized the similarity to Johannson's voice, after checking to be sure that OpenAIs process was 'clean', he should have shut up. Now he's sounding a lot like the Bart Simpson defense [wikipedia.org].

    • Why was he contacting Johansson? He was so hands off he tried to personally do licensing negotiations. Altman should have let sleeping dogs lie.

      He wanted Johansson because she would be a major PR win. He negotiated personally because she is a major star, doing so shows respect and make success more likely.

      Johansson did not want the role because of the likely embarrassing PR that would develop as the AI said embarrassing or offensive things.

      • 2 days before launch?

        • by drnb ( 2434720 )

          2 days before launch?

          Why not? It takes less than a day to build a professional voice model and there is hardly a shortage of samples of her voice. Any quirks in the model, any additional fine tuning necessary would be a minor thing. A word gets mispronounced or something. The real hazard in all this is not the text to audio conversion, it is generating the text in the first place. That's where the real embarrassing or offensive things will come from.

          "Professional Voice Cloning involves training (fine-tuning) the model on lar

  • Remember all those "average human" portrait morphs that tended to look very attractive?

      I'm not sure what the right term is, but maybe she just has a very "normalized" or "smoothed" sounding voice. In other words, as with those portraits perhaps her voice sounds like what a computer created average would, and this in turn is considered pleasant.

    • by HiThere ( 15173 )

      Also, WRT the argument about music, musicians have to WORK to avoid matching things that have been copyrighted, and they don't always succeed. (This is partially because of the insane rules about what counts as copyright infringement, of course.)

  • PR to English translation: OpenAI will 100% be doing this, as soon as we think we can get away with it and expect a profit.

  • although the two share a breathiness and huskiness" You just contradicted yourself in the same sentence.

    • "Sky sounds nothing like Johansson, although the two share a breathiness and huskiness. In a statement from the Sky actress provided by her agent, she wrote that at times the backlash "feels personal being that it's just my natural voice and I've never been compared to her by the people who do know me closely.""

      So the people who know the actress say she doesn't sound like SJ, but the people who her Sky say it does.

      I sticking with the theory Altman & Co. ran the actresses voice through an SJ filter. In t

  • Celebrities and actors really need to get over themselves. Their talent was always expendable and they were paid too much for it.
    • You could say that about anyone in their income bracket.

      Anyway, it seems like OpenAI owns the voice now. What would they do to you, if you commercialized a bot called Scarla, sounding exactly like Sky?

    • by j-beda ( 85386 )

      Celebrities and actors really need to get over themselves. Their talent was always expendable and they were paid too much for it.

      Yeah, we need to ensure that only the owners of capital can make money in our economic system. And not just ANY owner of capital, but the BIG owners of capital. Why do the top 1% own 50% of the world's wealth, when it could be even more?

    • Exactly. Due to their undeserved valuation they just have a lot of clout. If the voice had sounded (somewhat!) like even some B TV show actor (still a minor celebrity), the whole thing would have been completely ignored.

      Scarlett Johansson isn't even particularly known for her voice. She's mostly just a pretty lady with decent acting skills, like a lot of actresses.

  • Maybe it's analgous to something Sam Altman said earlier this month on the All-In podcast. "Let's say we paid 10,000 musicians to create a bunch of music, just to make a great training set, where the music model could learn everything about song structure and what makes a good, catchy beat and everything else, and only trained on that... I was posing that as a thought experiment to musicians, and they were like, 'Well, I can't object to that on any principle basis at that point — and yet there's still

  • by PPH ( 736903 ) on Saturday May 25, 2024 @11:29AM (#64498363)

    ... just go with Fran Drescher, I'll never understand.

  • OpenAI should really have gone forward with using the "sky" voice, IF it comletely relied on that actress voice instead of Johansson. If they could prove it was the other actress, who has a similar voice, used for the voice, and not Johansson it would have been an important lawsuit. Now it seems that a famous actress can stop an unknown from using her natural voice because she sounds too much like her, and THAT of course is ridiculous.
  • Sam: 'Google is launching a voice interface Gemini Live just like ours, how can we stand out better? Should we hire Scarlett Johansson to do the voice?"

    ChatGPT5: That's a good idea Sam but you'll get way more exposure if you can generate some controversy. Use a voice similar to Scarlet's and then have her act outraged and threaten to sue. This will create drama and bounce back and forth in the press creating publicity that eclipses Google's announcement, even if it has a better product.
  • I mean , there is thousands of people that sound alike. Real life sample, in my calls at work, there are 2 women voices that I cant tell apart. They sounds insanely close! Imagine that one wanted to sue the other for confusing me.
    • Yeah, this whole thing feels pretty ridiculous. It also seems not altogether inappropriate for someone who jumped to conclusions to subsequently apologize.
  • At this stage, it does not matter that the voice is from a different voice actress than Scarlett Johansson.
    By alluding that it was her, they have infringed on her personality rights [wikipedia.org].

    She has the right not to work for or promote any product she wishes.

    There are preceding court cases. There may be more, but so far I've heard of:
    * An ad using a Barbra Streisand-impersonator singing a song that Streisand was known for. It did not matter that the song itself was a cover.
    â Ads with a Tom Waits-impersonator, s

  • The voice actress should sue Johansson now for using the legal system to prevent her from being able to fully monetize her own natural voice. After the Open AI thing, she might have a hard time finding work as a voice actress as her potential employer will likely fear getting sued by Johansson.
  • Is you can ask AI what fraudulent collateral needs to exist in order to cover the tracks of your crimes. Understanding that that has been common practice for a few years now is pretty important to the discussion.

  • Customer: "I'd like a Coke, please."

    Coca Cola company: "No. Or at least no to your current offer."

    Customer: "Oh well. One Pepsi, please." [This is happening, whether you like it or not.]

    PepsiCo: "Here you go."

    Coca Cola company: "I sue!!1"

  • If the intent was to sound like SJ, then it might not matter whether they copied, used, or trained on SJ's voice. What might matter for IP law is whether they are trying to piggyback on her brand.

"Everything should be made as simple as possible, but not simpler." -- Albert Einstein

Working...