Because they are using her likeness without her permission. If they just wanted a “similar” voice they could have hired one from the start. They wanted her voice and when she said “no” they used it anyway.
The “her” tweet makes it exceptionally clear that it was always intended to be Johanssen’s likeness, not some other voice actor’s that happens to sound similar; OpenAI is just betting they can get away with it.
It seems like OpenAI were also betting on Johanssen not knowing her rights _de jure_in that area, which it appears they were quite wrong about. (And even if she didn’t, she would have quickly learned of them in this hyper-connected world, making the bet utterly incompetent…)
Because she inherently owns the rights to her voice and they’re attempting to reproduce it without her permission. This has been litigated in multiple cases. If they had just wanted to be _reminiscent of_ “Her” they could have a different outcome than if they tried to get the original “Her” voice actress and then tried to reproduce her voice in her she said no.
Intent matters a lot in a situation like this and OpenAI very clearly documented their intent for every me to point and laugh at. Even in the hypothetical case I give above where they don’t try to hire Johanssen and just get someone similar-sounding, they could still be liable for appropriating Johanssen’s voice if they were coaching the voice actor to specifically sound like Johanssen rather than do her own performance.
You’re looking for a technicality but this stuff relies on human judgment for a reason.
So based on that human judgment, no "similar" voice to a public figure can be used in a product/service/demo/ad etc. without that person's consent. What's next? Banning similar people from appearing in demos? Banning particular musical notes from being present in songs because someone else played that note in a song?
As long as it's not literally her voice (or a model trained directly on her voice) they should have rights to do anything they want.
Again, you’re looking for some sort of absolute that either allows or prohibits something in all cases based on a set of ironclad rules, and that doesn’t exist because intent matters in human endeavor.
If a similar voice was used because it was _reminiscent of_ “Her” but was still a unique performance, then it probably wouldn’t be a problem. Where there’s a real problem is if it’s an intentional mimicry of a specific person’s specific vocal performance.
And Johanssen can show OpenAI had such intent in court pretty easily, between their attempts to get her consent to use her voice (which she denied) as well as their CEO’s “her” tweet promoting the voice.
It’s extremely clear that their intent was to use her voice, not just a gestalt or tone, given that they approached her multiple times—including after they’d actually implemented it using another voice actor.
Do you really think that everything exists in a a vacuum and must be decided from first principles, giving the maximum benefit of the doubt to one party over another? There is extensive case law on exactly this situation, and OpenAI has run afoul of it.
No, if you happen to sound like a celebrity there's nothing stopping you from making your own way in the acting world. You should be careful not do anything that looks like you are deliberately trying to pass yourself off as that celebrity or profit from their persona, though.
Like, for example, voicing an AI chatbot when that celebrity, one of the most highly paid in the world, has previously starred in a famous movie as an AI chatbot. And especially when the makers of the chatbot have repeatedly publicly referenced the movie as inspiration in creating the chatbot and have publicly pointed out the similarity between their chatbot and the movie. And especially when the makers of the chatbot have approached that celebrity to ask to use her persona and been denied.
Because if you do that you might just get your arse sued off.
The “her” tweet makes it exceptionally clear that it was always intended to be Johanssen’s likeness, not some other voice actor’s that happens to sound similar; OpenAI is just betting they can get away with it.
It seems like OpenAI were also betting on Johanssen not knowing her rights _de jure_in that area, which it appears they were quite wrong about. (And even if she didn’t, she would have quickly learned of them in this hyper-connected world, making the bet utterly incompetent…)