OpenAI has decided to suspend the activation of one of the five distinct voices on its ChatGPT platform, after American actress Scarlett Johansson accused the company of intentionally copying her voice without her knowledge. The voice "Sky" is the one affected by this decision, as it closely resembles Johansson's voice, who played a voice assistant powered by artificial intelligence in the well-known film "Her."
In a post on its website, the company clarified that the selection of the five voices came after extensive testing by a panel of acoustic experts and a large number of specialists in selecting voice actors, based on performance and voice quality. OpenAI stated on Monday via its platform "X": "We’ve heard questions about how we chose the voices in ChatGPT," adding, "We are working to pause the use of Sky while we address them."
Despite the company confirming that the voice was developed based on various actresses’ voices, Johansson accuses OpenAI and its CEO Sam Altman of intentionally replicating her voice without her consent, prompting her to hire a lawyer to push for a change in the voice. The American star commented in a statement released on Monday: "Last September, I received an offer from Sam Altman, who wanted to select me to be the voice of the ChatGPT system." She added, "He said he believed my voice would comfort people,” but she confirmed that she "declined the offer."
Johansson continued: "When I heard the demo, I was shocked, angry, and in disbelief; Altman developed a voice that resembles mine so closely that my close friends and the media couldn't detect any difference," emphasizing that "Altman implied the similarity was intentional." She explained that she was then compelled to hire legal counsel who sent two letters to Altman and OpenAI, resulting in the company reluctantly agreeing to remove the Sky voice.
She noted, "As we all grapple with deepfakes and work to protect our image, work, and identity, I think these questions deserve absolute clarity," stating that she is "eagerly awaiting" appropriate legislation to help ensure the protection of individual rights.
The controversy over this voice has sparked widespread debate since OpenAI's conference to launch the new intelligent model GPT-4o, where the smart platform demonstrated outstanding voice performance, relying on "Sky." The voice capabilities with the new model succeeded in delivering communication closer to "human," accurately demonstrating a remarkable state of sentience.
In its statement, the company explained that the five voices offered for users to choose from in audio interactions with ChatGPT were selected from over 400 voices presented by many actors during a testing period conducted in May 2023, after the introduction of the Voice Mode feature in September of the previous year.
OpenAI outlined the criteria for choosing the ChatGPT voices, including that the actors come from diverse backgrounds or can speak different languages, that their voices are distinctive, do not reveal the actor's age, and are trusted by users, possessing warm, rich, and engaging tones, while being highly natural.
By June and July 2023, the specialized committee chose a set of voices, allowing the artists to move to San Francisco to begin professional recording trials, resulting in the selection of 14 voices, from which only 5 were ultimately chosen: Breeze, Cove, Ember, Juniper, and Sky. These names were chosen by the company for the voices and do not represent the names of the original performers, whose identities were kept confidential.
Subscribers to the ChatGPT Plus version are expected to use the developed Voice Mode with the new generation GPT-4o within weeks, with the same feature later available to users of the free version of the platform.
