OpenAI has announced in a message on X that “they are working to pause the use of Sky.” Sky is the voice she remembered, in the new ChatGPT, Scarlett Johansson’s tone in the movie Her, where a man falls in love with his virtual assistant. The message, published early in the morning in California, includes a post about how they have chosen and trained the voices of actors for ChatGPT. In that statement there is a surprising paragraph where the company specifically comments on Johansson’s case: “We believe that AI voices should not deliberately imitate the particular voice of a celebrity. Sky’s voice is not an imitation of Scarlett Johansson, but rather belongs to a different professional actress who uses her own natural voice. To protect her privacy, we cannot share the names of our voice talents.”
But on Monday afternoon — also California time — some of the reasons for this quarantine were known. Those who speculated on social networks that the pause came due to a lawsuit by the Oscar-nominated actress were not wrong. There has been no lawsuit, but there have been legal warnings, as she herself has said. Johansson published a statement in which she explained that Sam Altman, executive director of ChatGPT, had indeed offered her last September to hire her so that her voice would be part of the new ChatGPT 4.0. “She told me that she thought my voice could comfort people,” reads the actress’ statement, broadcast on X by journalist Yashar Ali and confirmed by the i’s publicists. “After long considerations and for personal reasons, I declined the offer. Nine months later, friends, family and the general public have been able to see how the new system called Sky sounds just like me. When I heard the demo, I was shockangry and incredulous that Altman had sought out a voice that sounded so eerily similar to mine that my closest friends and the media couldn’t tell the difference.”
Although OpenAI has allowed conversations with voices since September 2023, the hype with Sky’s voice only came with the presentation last Monday of its new model, ChatGPT-40, which reduces latency and allows conversations in real time. During the presentation there was a moment where the model seemed to flirt with company employees. Sam Altman himself published the brief message “her” on X during the event. The parallels were neither coincidence nor exaggeration. The company itself promoted them. Other employees released messages about the film. The actress also talked about it in her statement: “Altman even hinted that the resemblance was intentional, tweeting the word ‘her’, a reference to the film in which she played the voice of a chat system, Samantha, who creates a relationship. intimate with a human.”
The protagonist of titles like Lost in Translation y The girl of the pearl He continues to explain this and reveals that his lawyers gave Altman a notice that made him back down: “Two days before the launch of the ChatGPT 4.0 demo, Altman contacted my agent, asking me to reconsider. Before we could speak, the system was already in the air. As a result of these actions, I was forced to hire legal counsel who wrote two letters to Mr. Altman and OpenAl, setting out what they had done and asking them to detail the exact process by which they created the Sky voice. As a result, OpenAl reluctantly agreed to remove the Sky voice.”
Johansson continues in his writing: “In a time when we are all faced with the deepfakes and the protection of our own image, our work and our identity, I believe that these are issues that deserve absolute clarity. I hope for a resolution with the greatest transparency, as well as the approval of appropriate legislation that helps guarantee the protection of individual rights.”
We’ve heard questions about how we chose the voices in ChatGPT, especially Sky. We are working to pause the use of Sky while we address them.
Read more about how we chose these voices: https://t.co/R8wwZjU36L
— OpenAI (@OpenAI) May 20, 2024
For its part, OpenAI explains in its statement that they began looking for actors and actresses for their voices at the beginning of 2023. The personality of those chosen had to include at least all of these traits: “Different origins or those who could speak several languages, a voice that seems timeless, that inspires confidence, that is warm, attractive, that is charismatic, with a rich tone and easy to listen to.”
The company received 400 proposals. The candidates had to record several phrases and conversations and it was explained to them exactly what they were for. They also explain that the remuneration was high: “Each actor receives compensation above market rates, and will continue to do so as long as their voices are used in our products.”
Last summer they recorded their voices in person in San Francisco for ChatGPT. Looking ahead to the new model presented last week, the five chosen voices continued collaborating with new material. The statement does not give any other details of what will happen with the Sky voice, beyond that there will be “more options”: “Looking ahead, there will be even more options, as we plan to introduce new voices in ChatGPT to better adapt to the various interests and preferences of users.