Tom Waits vs. Frito-Lay, Inc (2003)
tomwaitslibrary.info
tomwaitslibrary.info
My very favourite line, among many, from "Step Right Up": __The large print giveth and the small print taketh away.__
https://en.wikipedia.org/wiki/Anywhere_I_Lay_My_Head
Eclectic choices even for Waits songs. Though I'd like to hear her take on Poor Edward or Fish And Bird or Soldier's Things
I think that's why this was posted... I saw it referenced in a comment on that thread.
Coevolution. The AI and the now-famous actor are the end products of a selection process aimed at a similar audience. The AI creators were likely aiming for the most attractive female voice, according to a bunch of young men. That parallels the process used by those casting action movies, which are also aimed at young men. It should be to nobody's great surprise that the end products of both share many features.
What is very odd is how similar celeb faces actually are. There really is a mathematical standard for beauty when it comes to female faces, which is why so many celebs are have eerily identical proportions. There is probably a model too for voices.
https://gotobeauty.com/plastic-surgery/how-math-shapes-moder...
This might hold water if the actor weren't famous for more than a decade before the voice model existed
"Passing similarity"? All her friends and coworkers say it sounds like her - she describes being contacted by friends who thought she'd done it. A lot of the public, including HNers, identified the voice as hers, independently - not in some A/B test, but they heard it and said to themselves "Johanasson did the voice for this."
You're the only person in the room who thinks OpenAI didn't use samples of her voice for their voice model - or that the voice has "passing similarity."
Altman approached her asking her to do it. She refused; they went ahead and did it anyway, probably using interviews since the audio would be very clean - and two days before they went live with it, Altman tried to negotiate a second time and was rebuffed. Released it anyway.
Why would Altman feel the need to panic-negotiate a second time, days before they went live, if they hadn't used her voice for it?
If OpenAI designed it to sound like her, that's Midler tort. They actually used her voice to train it, which means Midler tort and more. The question will be how much more - I hope they get sued into the ground.
Really? No one seems to be claiming that anywhere I've seen.
We know that they really did use a voice actress, and that it really is her actual voice, and that she was hired for it before ScarJo was approached. There's no "panic negotiation" there -- the most panic needed would be if someone internally identified that there could be a PR issue if people mistakenly thought it was close enough to seem like intentional impersonation.
Of course they wanted ScarJo, once they thought of it! It'd be great publicity. But that doesn't then mean that anyone who sounds somewhat like ScarJo (but certainly not identical) becomes retroactively unusable or makes it vocal impersonation.
yes. yes it does. That's what TFA is saying. Waits' case is exactly about this. They hired a singer to make a recording to sound like Waits. OpenAI hired a voice actress with a voice like Her to make their AI sound like Her. And then their CEO tweeted "her". It's that simple.
OpenAI hired a voice actress with a pleasant, low, breathy voice who shares a hint of a regional accent with ScarJo to be one of a bunch of voices for their models. Eight months after releasing it, the CEO tweeted "her." It's not illegal to hire someone who sounds a bit like a celebrity for something, and certainly not one who literally sounds more like a different celebrity (Rashida Jones).
This isn't a violation of rights of publicity for a number of reasons, but the most dispositive of them is that she doesn't actually sound that much like ScarJo. There's no vocal fry! That's the most distinctive part of her voice!
It takes a while to get a lawsuit going. According to the article, Waits learned about the ad on October 3, 1988, filed a suit in November 1988, and the case was tried before a jury in April and May 1990.
With sj, they approached her 6 months ago, she declined, and again just prior to launch to which she didn't respond. And Sam tweets one word 'her'.
This is a slam dunk and they will lose.
Her agent, who has requested anonymity...
Tells WaPo that his client, who also "needs to remain anonymous" due to "fears for her safety" tha she told him that OAI "didn't mention SJ".
Tells WaPo that she gave him a statement that they can print...
What are these fears for the actress' safety that require her, and her agent, to remain anonymous?
Why was OAI ready with a trove of legal-reviewed documents to hand to WaPo the moment this blew up?
Willie Williams https://www.youtube.com/watch?v=G1KV51qIdI0
Howlin Wolf https://www.youtube.com/watch?v=9Ri7TcukAJ8
or Screamin Jay Hawkins https://www.youtube.com/watch?v=7kGPhpvqtOc
It surprises me a little because the Dorito commercial sounds like a fairly generic blues backing track and a raspy voiced singer. From the article, there's an indication that Doritos was looking to replicate the sound and feel of the song, and they were aware of the legal concern. But to me it kind of sounds like Tom Waits is doing an impersonation, the Doritos song is doing an impersonation, but it's not totally obvious from the song that the Doritos impersonation necessarily goes via Tom Waits rather than back to the original source material.
Like Plastic Factory https://www.youtube.com/watch?v=soV8IcTzuj8
The commercial was using a tom waits song sung by an artist who impersonated Tom waits and the folks involved discussed the potential legal implications and did it anyways.
See e.g. https://www.youtube.com/watch?v=iG0p2JHcsQ0 . It's playful, original, pulls from a lot of previous artistic work.
Later on he does the soundtrack for The Black Rider, changes record label and gets the freedom to dabble in post-punk and weirder stuff.
Change your shorts, change your life
Change your life, change into a 9 year old Hindu boy and get rid of your wife
I think nowadays, the idea that doing a commercial damages your artistic integrity is considered sort of an anachronism. For better and worse, I guess, and largely driven by the consequences of technology in any case. But, a similar lawsuit today would probably be about damaging the monetary value of the artist's brand, rather than about damaging their artistic credibility per se, as it was with Waits at the time.
It used to be a truism that taking money from sponsors changed your allegiances, and changed the nature of your art. For the worse. Whether time has been proven that true or false is an exercise for the reader. Point is, that's where Waits seems to be coming from. It's not about "they didn't pay me first," it's about the commercial diminishing his credibility because it looks like he might have agreed to do it in the first place.
I'm not sure this is true. It was always the case that these things were contextual and based a lot on how the artist/musician/author presented themselves. The whole idea of "selling out" is predicated on the assumption you have some high ground to give up, or at least that the action you are taking would undermine strongly held beliefs of your fan base.
When popularity swings more to (obviously) commercially driven output, there is plenty of selling, but less selling "out", as it were. The pendulum will likely swing again.
Personally, I think it's more a change in how artists view themselves in relation to the market—and how audiences view them as well—and that artists can still stay true to that understanding and have it be a kind of ethos... just a very different one than we had when I formed my opinions on music.
What I'm saying is, artists can still perform acts of betrayal, and lose their credibility to their audience. So, they must have some to lose it. It's more that licensing their music is not one of those acts anymore.
ska can correct me if I'm wrong, but I think what ska is saying is that the focus of the population is currently on artists who don't value their authenticity.
There remain many artists who are associated with authenticity, but they've lost the focus of the general population (for now).
There are certainly some genres where the audience values this sense of authenticity of course. That and people like Waits who actively cultivate this as part of their public image. Otherwise I wouldn't expect it as a norm.
Personally, I'm still waiting to see Maynard sing Hooker With a Penis on a Coke commercial.
Being a successful businesswoman is part of Taylor Swift's artistic identity, where it wasn't for Loretta Lynn. Selling sneakers is part of Kanye West's artistic identity, where it wasn't for Tupac. It doesn't make any of them less authentic.
https://www.wsj.com/articles/these-are-the-fashion-brands-ra...
Meanwhile, I can't imagine any band in the punk or metal genres making it very far if they leaned into commercialism. That would not be authentic at all.
Here’s a clip with the part that comes after the ad:
Scarlett Johansson might beg to differ.
I think some people just happen to create art.
I think others make it part of their personal definition, and to those integrity is part of it (if they choose to have it)
Is there any reason why Elektra/Asylum didn't license the actual recording copyright[0] to Frito-Lay? I'm assuming Tom Waits (like any other musical artist) wouldn't have veto rights over licensing the recordings, in the same way he apparently couldn't stop Fifth Floor Music from licensing the song itself to Frito-Lay.
The thing is, if Frito-Lay had actually licensed the recording, Tom Waits wouldn't have a leg to stand on in court, because of a very funny concept in copyright law called federal preemption. Any claim under any other law - state[1] or federal - that looks and quacks like a copyright is null and void. You only get to sue for copyright with copyright. So you can't, say, trademark a public domain work[2] and then sue people for reproducing it. Misappropriation, false endorsement and publicity rights are very much trademark-shaped laws, so they also lack any jurisdiction over copyright matters. There really just isn't room in the law for "I license you this work" but also "you reproducing this work is a false endorsement". The public is not confused when copyrighted works are used with permission.
However, I'm also not sure why suing for copyright infringement was off the table in the first place. The thing is, when you make a derivative work, you own what you added. If you and me both go to Disney and buy licenses to produce Avengers merch, but I decide to copy your design for the merch, you get to sue me. My license to make my own derivative version of something does not entitle me to copy other derivatives of that same work. So Frito-Lay, having a license to record their own version of Step Right Up, doesn't get the right to copy Tom Waits' recording of Step Right Up.
Who knows, maybe recording copyright is a lot narrower than other forms of copyright, but it's hard not to shake the feeling that he could have gone up against Frito-Lay for a lot more.
[0] Music copyright has two souls: the copyright over the song itself - lyrics, sheet music, and so on - and a separate copyright over a recording of a specific performance of the song. Originally you could only copyright the song and not the recording.
[1] The reason why federal preemption exists is that states started inventing their own recording copyrights for music. Which sounds absolutely wild to lawyers today, who are taught that copyright is inherently a federal question and that states have no say in how it works. What's even more wild is that some state recording copyright laws were actually perpetual, this somehow survived the "for limited times" language in the Copyright Clause, we didn't establish federal preemption and shut down these schemes until the 1970s, and we didn't extinguish already extant perpetual recording copyrights until the Music Modernization act in 2018.
[2] i.e. Disney putting Steamboat Willie in their logo
Eg:
That's right, it fillets, it chops
It dices, slices, never stops
Lasts a lifetime, mows your lawn
And it picks up the kids from school
It gets rid of unwanted facial hair
It gets rid of embarrassing age spots
It delivers a pizza
And it lengthens, and it strengthens
And it finds that slipper that's been at large
Under the chaise longe for several weeks
And it plays a mean rhythm master
It makes excuses for unwanted lipstick on your collar
And it's only a dollar, step right up
It's only a dollar, step right up
A big part of that era’s music industry was being able to convince musicians they weren’t totally fucked as soon as they signed.
> When I was a kid, if I saw an artist I admired doing a commercial, I'd think, "Too bad, he must really need the money." But now it's so pervasive. It's a virus. Artists are lining up to do ads. The money and exposure are too tantalizing for most artists to decline. Corporations are hoping to hijack a culture's memories for their product. They want an artist's audience, credibility, good will and all the energy the songs have gathered as well as given over the years. They suck the life and meaning from the songs and impregnate them with promises of a better life with their product.
Tom Waits did have a point that I think today's content creators need to take onboard. With music it was not always about money for everyone, the love of music was motivating enough, bringing people together for a good time.
I do not see many content creators in it 'for the content' and bringing a community together. There are definitely some but the algorithm isn't helping them.
> Eventually, artists will be going onstage like race-car drivers covered in hundreds of logos. John, stay pure. Your credibility, your integrity and your honor are things no company should be able to buy.
I wish politicians were obliged to wear suits decorated in all the logos of their sponsors.
Not so coldly charted, it's really just a question of your honesty
It doesn't mean that that kind of music isn't any good, but it's often entirely commercial from the very start which makes it hard for artists who had little if anything to do with the creation of those songs to care about "selling out".
If you mean 60s and 70s artists cared about their art, whereas now they primarily care about the money, then yes.
>Like, today, most rap songs with explicit lyrics have a sanitized version for general broadcast. Music artists in the 60's would have fought hard against that.
Fighting for something you believe in and/or the integrity of your song? Such suckers!
Most rap songs are explicitly bragging about how they will do anything to make money, and about how people that won’t aren’t morally superior but are either born privileged or just making excuses for being weak and soft.
It’s alien coming from an upper middle class white family, where people try to pretend they have less money and care about money less than they actually do, because it isn’t socially acceptable- people hide their supercar in the garage and drive an old Toyota when people they know are looking. These people are just as ruthlessly greedy as Biggie claims to be, they just hide it.
The difference is that artists used to be able to earn a decent living from selling their art.
Today, artists (unless they are actually producing physical artifacts) are expected to give their work away for free (or for so little as to be pointless).
A true artist. Here's my favorite work of his: https://www.youtube.com/watch?v=Psk3rmjonQA
There was a time in my life when I was aware of just about every bit of Waits arcana available to the public.
Then I got older and lost my way.
Thank you for that link. #PEHDTSCKJMBA
https://www.theguardian.com/music/2016/oct/08/tom-waits-angr...
Case in point, Owen Wilson was the voice of the main character in the Cars movie, but they got a different voice actor for some of the cartoons who sounded similar. Same thing for the ghostbuster cartoon after the movie was a big success in the 80s.
Why is that ok?
Reminds me of an Archer bit: https://youtu.be/c9uuITbtl-g?t=2
- Oh my god, Slim Goodbody!
- No! No, this is absolutely not that trademark character. Just a unitard with the systems of the human body on it. On a guy.
- On a guy named TV's Michael Gray.
> All credits to FX. I do not own any rights.
Appropriate
Additionally there is usually a provision explicitly allowing for this in the contract with the initial voice actor. Depending on the voice actor’s leverage, they can negotiate for things like approval over the replacement, right of first refusal to voice the character or payment when a sound-a-like is used.
The fact the character sounds exactly like Owen Wilson himself is somewhat incidental though understandably makes this confusing. What they couldn’t do in this case is have an Owen Wilson soundalike voice a DIFFERENT Disney character. They only own Owen Wilson’s voice as it pertains to portraying the character Lightning McQueen.
In the OpenAI situation and Frito Lay there is no initial contract granting any rights to a voice performance.
What if she wasn't doing her own voice? Many actors have voiced characters not in their own voice over the years. Does Elmo's original voice actor own Elmo's voice or is it whoever owns Sesame Street?
I'm not necessarily making any assertions, I'm genuinely asking cause I don't know what kind of precedent is here. Though personally I'm not convinced of the case against OpenAI, other than bad optics from Altman.
That is likely the case they would try to make if they went to court. But this likely will be settled out of court if there's anything there
OpenAI trying to contract Scarlett twice would likely put a big damper on the "We didn't want it to sound like her, we wanted it to sound like the character" too
It's definitely not cut and dry
I think they 1000% did, because despite all the fuss there is actually no resemblance between the Sky voice and Scarlett Johansson's. The Sky voice is in a different pitch, doesn't have the vocal fry, doesn't have the slightly nasal tone. Listen for yourself[1].
This was a marketing coup for both OpenAI and Johannson. OpenAI, because they're in no danger of a negative judgment, Johansson because she gets her name all the place in the press and drums up free publicity for her film.
[1] https://www.reddit.com/r/singularity/comments/1cx1np4/voice_...
In the sample you linked to you're correct, but in this sample the vocal fry is present and in fact quite pronounced:
https://m.youtube.com/watch?v=vgYi3Wr7v_g
I can't hear the nasal tone in either voice, but the video linked above sounds much closer to Johansson's than the sample the reddit poster chose. It's definitely not a home run, though.
Cynical.
There's nothing kind about man
You can drive out nature with a pitch fork
But it always comes roaring back again
For want of a bird
The sky was lost
For want of a nail
A shoe was lost
For want of a life
A knife was lost
For want of a toy
A child was lost"
I like that Waits never became an entertainment industry pushover.
Never attribute to malice what can adequately be explained by incompetence (at least, until the legal team gets involved).
But in my experience, this is how 95% of corporate controversies happen - just some guy somewhere is an idiot and no one checks their work. But as soon as lawyers are involved they will pull out the Magna Carta if they have to in order to prove that you they were actually geniuses and within their rights to screw up.
Oops. I hope they get that fixed. I love Tom Waits.
That said, this is certainly a problem we will see in the future.
https://www.washingtonpost.com/technology/2024/05/22/openai-...
Copying a comment from another HN user (not sure if appropriate to name them):
* This one: https://www.youtube.com/watch?v=vgYi3Wr7v_g
* compare it to: https://youtu.be/GV01B5kVsC0?feature=shared&t=158
That seems like quite a strong resemblance to me.
I wanted to laugh at and belittle openai when I first saw scarjos claim that her friends and family couldn't tell the difference but from the samples you just linked I find that so unlikely that I think she is lying.
> It looks? like you're part of some recording? or production setup? With those lights? tripod? and possibly a mic? It looks like you might be gearing up to shoot a video? or maybe even a live stream?
They both have a vocal fry, but the former sounds more like valley girl without the slang or vowel lengthening.
The problem OpenAI has is they tried to get Scarlet to do the voice and after she declined OpenAI found someone who has a similar voice to do it. That shows intent which would otherwise be difficult, combined with the actual result being closer to Scarlet’s voice than Sky is naturally and they are likely going to lose.
Edit: By after I am saying they decided to use Sky’s voice not that it was recorded first.
As the public currently knows based on reporting from WaPo, the original VA was selected and performed her role (months) BEFORE they reached out to Scarlett for the first time. Please don't misconstrue the timeline for other people that are just now tuning in.
The core question eventually put before a jury would be if it was unintentionally similar or an imitation and attempting to pay her for use of her voice is clear evidence they wanted it because of it or they perceived the voices where overly similar. Either of those are possible, but to then use the voice anyway became problematic.
I’m not saying Sky did anything wrong. But the final result is close enough many people assumed it was Scarlet without promoting, that’s problematic when they are making references to her role as an AI voice and obviously wanted to make a deal.
And of course, some specific aspects of the case are unique. Universal used a face mold they had made of Glover's face, for instance.
I’ll give you the mold makes the intent more clear, but I doubt the jury is going to be debating intent. Pulling the voice suggests OpenAI thinks this is either a real risk or bad publicity.
Really using a voice on a digital assistant that sounds like the actress playing a role of a digital assistant is just dumb. Especially if you then tweet about the film days before releasing the voice.
Here’s what probably happened. OpenAI people realize they can do TTS for ChatGPT so they hire some voice actors. Some people point out that it would be cool if it could be as good as it is in Her, especially with 4o on the way, so they decide to contact Scarlett Johansson either to do a voice or just some promotional thing. She declines, but after 4o is released OpenAI realizes they can fairly make the comparison to the movie based on the quality and speed of the voice. Simply subtly referencing the film via an obvious parallel doesn’t constitute deceptive impersonation.
Personally there’s many ways they could have been less suspicious from my perspective. Have a large number of voices or the ability to tweet them in many ways and no one voice seems important. Contact many famous people and it would have seemed like a promotional rather than a reference. Have someone lower on the totem pole than the CEO of a multi billionaire dollar corporation trying to make the deal. This not being about an actor playing an AI assistant in a movie vs the voice of an AI product. Not actually making the reference in a tweet.
The phrase ‘the preponderance of the evidence’ comes to mind.
There’s a chance they did contact other celebrities. We only know for sure they contacted Scarlett due to her lawsuit.
Here’s the thing. No matter what, people would make the reference to Her. It’s like how everyone would make the connection to any sci fi movie if a device came out similar to the one in it. A woman’s voice as a virtual assistant is a very broad UI category, and being cheeky and drawing the comparison to a specific movie on everyone’s mind doesn’t constitute misleading impersonation, especially when you consider from OpenAI’s view the fact that they hired a different actress and didn’t ask her to do a Her impersonation.
> especially when you consider from OpenAI’s view the fact that they hired a different actress and didn’t ask her to do a Her impersonation.
That’s not specifically what they are accused of doing. The voice actress’s raw audio is only half the story, how they manipulated that audio makes a real difference in what the model sounded like. As many people assumed it was her objectively the end result was close. Worse the model sounds closer to Scarlet than the voice actress does.
Put another way, they wanted to end up at X and also happened to end up close to X when the search space was huge. Where did you get that Huge TV? It’s the craziest thing it just happened to fall off a truck.
If they had released the voice they could have argued they wanted something different, but by waiting and going with Sky after the negotiations failed it’s a different matter.
they tried to get Scarlet to do the voice and after she declined OpenAI found someone
This is exactly the question the Washington Post set out to answer, and the answer is "no": that didn't happen.
Simply recording Sky’s voice beforehand doesn’t imply a different intent. This isn’t animation where you’re trying to match two mediums and you want the actors recording first. Having a similar voice to work with would have been useful internally to know if they wanted to pay what it might cost to use her voice.
But it’s a small number of voices who all sound distinct. So it seems unlikely they wanted to released two similar voices in that context. Especially when it specifically sounds like the voice actor in a film as an AI.
They still created a voice that sounds like ScarJo - and many signs point to that being intentional. Overall, not unlike the case in "Tom Waits vs. Frito-Lay, Inc".
A voice isn't that unique in the grand scheme of things.
https://www.vice.com/en/article/wxepzw/musicians-algorithmic...
The Waits case is highly dissimilar, IMO. But tbf this isn't an area of law I'm expert in, and my knowledge of Right of Publicity law is mostly from law school, which was... a while back.
It, at best, makes it slightly less likely that the original intent was commercial imitation, but it doesn’t do anything to refute the case that that’s what the ultimate use was. Adding one extra round to the back and forth of how they got to that point doesn’t really change anything important.
You seem to be looking at this as if right of publicity were like copyright, where what is essentially protected is, well, the act of copying.
Right of publicity is a separate area of law, but it is more like trademark than copyright. The mechanism by which a resemblance which is commercially leveraged is attained is not relevant.
The Midler case was a vocal impersonation using one of Midler's songs. As discussed in the Midler case, where there's an imitation of the voice, one of the keys is also how much of the likeness/recognition/brand of the person is tied up in their voice, and that's substantially different between a singer's voice on their own song vs. an actress's voice. And given that it's not an impression/impersonation, and literally sounds more like another famous actress than it does ScarJo? This is a great example of facts you'd give to show where the Midler precedent doesn't apply — it's certainly in no way "exactly" like Midler.
Would you be willing to bet a small sum on this?
All SJ has to do is show that they wanted her and plant the idea that they may have been thinking of her when they hired the other actor. The end results sounds enough like her that I think OpenAI is going to have a very rough time in court.
I think the chances of this getting to court are almost nothing. OpenAI will replace the voice and settle with SJ.
Remember the Blurred Lines vs Got to Give It Up lawsuit? I personally think they sound way different yet Robin Thicke and Pharrell Williams lost that one.
Wait, are you alleging that an individual has a right to faithful imitations of their voice, as well as to their actual performance?
So if I do a really good Morgan Freeman impression and I want to use that voice for a voiceover that I make, and I intentionally do an impression, he should have the right to tell me I cannot use the voiceover that I made, because I intended it to sound like Mr Freeman? Even if there is no intent to mislead and it's rightfully credited etc (not like the Tom Waits situation in TFA where the ad was intended to deceive)
What stops that from extending to parody? Should Sarah Palin be able to sue Tina Fey for the impression on SNL all those years ago?
If all it boils down to is >The end results sounds enough like her
this starts to sound like celebrities with sufficient popularity and clout can trademark likenesses that are sufficiently similar to them. So is the voice actor who happens to sound like Scarlett Johnansson but is insufficiently famous not allowed to use their own voice in recordings because it might sound too much like the famous person?
Absurd absurd absurd
Nobody would confuse Tina Fey for Sarah Palin, while it would be possible for someone hearing the "Sky" voice combined with the "Her" aside to assume it was SJ providing the voice.
Again it would be up to a jury to decide, but there are plenty of previous cases decided in the artists favor that any sane lawyer would be uneasy taking it in front of a jury.
Surprisingly many people think "I can see Russia from my house" is a thing the real Palin said.
That's where this has always gotten odd to me. There's obvious impersonation, sure, but what if this same actress wanted to e.g. be the Major in a new Ghost in the Shell dub? Is she responsible for making sure everyone who hires her very definitely isn't hiring her as a soundalike if she wants her work to get released?
I'm sure it would be fine in reality, but if saying "her" (when advertising an app you can have a conversation with) is enough to make this impersonation, then the same logic says she's barred from certain roles (or categories of role) because someone more famous got there first.
I’m thinking of a number, bigger, no smaller. Keep playing the game and the end result is arbitrarily close to the number you’re thinking of.
The search space for generic seductive female voice is huge and they happened to end up with one that objectively sounds like someone they tried to recruit. They can’t argue random chance at this point.
Of course they didn't mention "sound like SJ" in the casting call - it might as well say "Please violate Midler tort." Actors wouldn't do it, casting agencies wouldn't do it. She / her agency would have found out almost immediately, and C&D'd them into the ground within hours.
I'm not sure why you're believing OpenAI when they say that the voice actor is who they actually used, and not that after bringing in the voice actor, they didn't just toss the recordings aside, have an intern collect clips of SJ in interviews, and throw that in to the machine?
Altman has a long history of demonstrating incredibly poor morals. Stop taking anything he says at face value.
Yeah, not really damning evidence either way but it certainly looks like, according to their own words, OpenAI intended to use her as marketing.
He was probably not personally aware she was contacted, and it just so happened that his dumb movie reference hit a nerve.
The alternative is that he knew they were ripping off Scarlett Johansson and then publicly compared them. Which seems too stupid to be true.
People, even Silicon Valley CEOs, doing things that seem “too stupid to be true”, happens a lot more than most people think.
OK, but then they openly blast it all over Twitter? If you’re so open about what you’re trying to do with everyone in the world, yet you didn’t bother mentioning to the voice actress who’s supposedly going to implement this?
I didn’t really think it sounded much like Johansson, but it looks like OpenAI decided to pull the voice(?). They are in a better position to evaluate their liabilities (or the PR cost of having the perception that are ripping her off, whether or not it is right). Hopefully they’ll add it back, Johansson doubles down, and then we can get a court case to satisfy everybody. Both entities have plenty of money to spend on this, having them battle it out seems like a real win. (It would have been a real shame if a small no-name voice actor without the cash to fight OpenAI felt copied instead).
The voice chat didn’t seem to be particularly well publicized. People mention how after the 4o announcement, many people started voice chatting for the first time (and mistaking it for 4o voice chat) because they hadn’t realized the voice chat was there before.
OpenAI might have had a leg to stand on if their CEO hadn't gone on the internet and blasted out "Her" when advertising the Sky voice, and another co-founder had not specifically name-dropped using SJ to market their voice product. (And reading the WaPo article, they hired a film director to handle the voice recording, and the individual in charge of artistic decision-making, apparently an avid Her fan, was conveniently not made available to interview...Hm...)
Recently I had a friend message me to ask me if it was me asking a question at a recorded event with audience questions that they were watching on YouTube (the camera only showed the stage, not audience members at the microphones).
They had absolutely no reason to think I would have been at that particular event.
And I told them, yup -- it was me asking that question a few years ago.
And there is nothing particularly distinctive about my voice. If anything, my voice is probably particularly generic. But it's just timbre, accent, and the unique "fingerprint" of my particular vocal personality.
How many hours has your friend heard you talk? How many hours has the average person heard Scarlett talk? Coupled with the fact that Scarlett is acting when people generally hear her voice, which means some aspects of her vocal qualities change(black widow doesn't sound exactly like her character from Her)
And obviously we're talking about "normal" voices as opposed to impressions or accents. But we still recognize voices even if the voice is doing a wide range of emotions. The same way we don't stop recognizing someone's face when they change from a smile to a frown.
I'm beginning to wonder if you've ever considered your interactions with other people, or if you're just making this up as you go along.
We certainly have memory, and can forget people, like a random person we were once shown a picture of. And pictures are not 3D faces either.
But if you meet someone for half an hour and then run into them later that day, you recognize them just as easily as you do your mother. There's zero doubt that they are the person you met earlier (unless you have face blindness or something). You don't have to meet them for 10 more hours to be able to recognize them instantly.
Sure you might forget them a year later but that has nothing to do with your recognition capabilities, but with memory.
Which, given the fact that the voice actor for Sky has put out a statement stating that her voice has never been compared to Scarlett's in real life, makes me think that people are just hallucinating that Sky sounds like Scarlett. For what it is worth, I listened to them side by side(before OpenAI put out a statement by the voice actor) and they are not the same voice.
https://en.wikipedia.org/wiki/Midler_v._Ford_Motor_Co.
(TLDR; Ford licenses a Bette Midler track and gets a sound-a-like to perform it for a commercial after Bette Midler refuses to participate. Midler sues and wins)
> In sum, our holding in Midler, upon which Waits' voice misappropriation claim rests, has not been eroded by subsequent authority.
eta: all the folks saying the voice actor doesn't sound exactly like ScarJo and was hired before the negotiations with ScarJo are neglecting to remember: AI is really good at making voice imitations from snippets. So Sky's voice is almost certainly not unprocessed voice acting. That's how the There I Ruined It guy does his covers.
Perhaps Andrey Karpathy if he has a change if heart - he's been hinted to have led the AI personality effort at OpenAI before he left.
If not him, then the discovery process ought to be good. Scarlett's lawyers asked for details on how OpenAI trained the voice, then the voice got pulled ¯\_(ツ)_/¯.
If/when the voice returns and this goes to court, we may find out if there was any fine-tuning on ScarJo's interviews/movie audio.
It probably could be stretched to musicians as well. Maybe lookalike actors too if we go down that route.
I just can't imagine it'd be a good thing to have your voice be owned by someone else just because they became famous first.
Waits won because they were clearly intending to impersonate his voice, while singing his song. Not because a random guy on the radio sounded like him.
Ditto for SJ--there was a clear intent to reproduce her specific voice, based on her role in "Her".
Again, there wasn't. An article literally came out today that they had contracted the VA before contacting SJ. [1]
Honestly this argument right here is what the slippery slope argument is all about. If you continue any further, you're just further proving it. They literally didn't even contact SJ before hand.
>while singing his song
You even added this caveat before, its already the slippery slope in action. No song was being sung here.
[1] https://www.washingtonpost.com/technology/2024/05/22/openai-...
I'm betting on the actor sounding nothing like SJ at all. Their voice was used as a baseline in the training of the tool, but ultimately it was modified to copy SJ.
The potential legal issue is much more about the details of the commercial presentation and marketing of the work by OpenAI than it is about her work itself, beyond whether the work has sufficient similarity to make the particular manner of commercial use a legal issue. There’s nothing for her to defend.
> I just can't imagine it'd be a good thing to have your voice be owned by someone else just because they became famous first.
Are you saying that anyone being able to train generative AI on someone's work or a facsimile to directly compete with them is actually better for commercial artists?
Also, AI to change my lip movement so that it does not look dubbed.
I think the Marvin Gaye ruling is concerning, especially because no mere mortal could hope to either launch or defend themselves against such an action. However, I don't think those extremes are a reason to just say "oh well. I guess fair use should apply to nearly all derivative use, even commercially. If that means creativity is only commercially viable for AI companies now, so be it." Professional creative expertise can't be replaced by generative AI, and it's important to or society, but copyright is the only set of guardrails on the only viable market for many creative fields. I'd love to abolish copyright, but first is love to live in a society that could support the millions of people that use it as currency.
Backstory: https://theguardian.com/technology/article/2024/may/20/chatg...
That seems to be a leap. It's not at all hard to define a bright line between two people that merely sound the same and inappropriate appropriation. Lots of folks have deep voices similar to James Earl Jones, but if you hire one of them to voice a helmetted character named Tarf Later you'd expect to lose a suit under this statue, right?
The case at hand isn't that Sky sounds like Johansson in the abstract (whose actual voice isn't even all that unique or notable), it's that explicitly evocative of her role in Her, and most damningly that they clearly tried to hire her to do it.
First, the "Her" reference from OpenAI.
Then ScarJo saying "Hey that sounds like me. Even some of my friends think so."
The voices sound kind of similar in some ways, and dissimilar in other ways. If the voice actor was trying to mimic ScarJo, she didn't do a very good job.
Is a casual reference to the title character exemplifying the same concept, a female-voiced AI, in a movie that won best screenplay at the Academy Awards and Golden Globes, an IP violation? Even if that were the case, it would be a studio matter and not Scarlett's IP.
I am curious why they reached out to ScarJo again 2 days prior, though.
Did they want to use her purely for marketing? That seems doubtful, because they'd have to get movie studio clearance to use "Her" in official marketing.
Did they have a separate model trained on her voice (I wouldn't put this past OpenAI) and were hoping they could get last-minute clearance to use it? This is actually my suspicion. That failed, so they just went with the voice models they already had clearance to use. That's not illegal. It's not even unethical. Anyone can try to train voice models on voice samples they collect. What's a problem is commercial use and representation of likeness.
I don't think a casual reference to Her was a representation that the voice is like ScarJo. It was merely referencing (very effectively) the concept of the movie's always-there [female] AI-voiced AI-chatbot.
It's been pushed down the Memory Hole.
My take is that for a certain cohort of AI hypersters and Sam fans, it is literally impossible for OpenAI to do wrong. Sam could probably shoot a man in cold blood on national television and we'd have somebody explaining how technically it's legal because reasons.
Any other examples? Not trying to defend OpenAI just interested in this idea that it has actually made anyones lives worse.
The article explicitly mentions the Tom Waits case did not include a copyright component as he didn't own the copyright, and Frito-Lay probably obtained synch rights from the copyright owner.
Tom Waits won the suit without invoking copyright violations - read into that what you may.
If OpenAI's lawyers thought they were going to get away with it then OpenAI needs to get better lawyers.
Part of being a good lawyer is sometimes "well, this didn't work in 1992 (or whatever) but it might fly now. Let's try it, worst case is X". And, as always, the devil is in the details.
This one is unique because Hauri (aka "Yung Gravy") is another music artist. The Midler case is actually mentioned in the court filings, too.
It's interesting to think about the blurry line between art and commercial products.
I wouldn't be surprised if there were literally thousands of voice actors that sounded close enough to any given actress to pass as her.
I'm pretty sure this is exactly why TFA was posted—apparently you can, in fact, protect a voice under California law. Of the $2.6 million in damages awarded in TFA $2 million were for "voice misappropriation".
This seems to be the relevant section of the civil code: