OpenAI has pulled one of the ChatGPT voices named "Sky" following controversy over its resemblance to the voice of Hollywood actress Scarlett Johansson in Her, a film in which a lonely man named Theodore (played by Joaquin Phoenix) falls in love with a virtual assistant named Samantha.
The 39-year-old actress’s legal team has sent OpenAI a letter asking the company to detail the process by which it developed the voice.
About a week ago, the company announced its new model, called GPT-4o (the o stands for "omni"), which allows ChatGPT to talk to users in a more lifelike way while detecting emotions in their voices, analysing facial expressions and changing its own tone depending on what a user wants. The company offered examples during a live event. Asked for a bedtime story, the voice was lowered and a playful tone was heard.
The company said the five voices — Breeze, Cove, Ember, Juniper and Sky — were selected through a casting and recording process over five months. There were 400-odd submissions from voice and screen actors and the figure was brought down to 14 and eventually the final five.
Scarlett Johansson said in a statement that last September she received an offer from Sam Altman to "voice the current ChatGPT 4.0 system".
"He told me that he felt that by my voicing the system, I could bridge the gap between tech companies and creatives and help consumers feel comfortable with the seismic shift concerning humans and AI. He said he felt that my voice would be comforting to people. After much consideration and for personal reasons, I declined the offer."
The Black Widow star said she was “shocked, angered and in disbelief” that Altman would “pursue a voice that sounded so eerily similar to mine”.
When Altman was asked about his favourite sci-fi film at a conference in September last year, he said: “I like Her. The things Her got right —like the whole interaction models of how people use AI — that was incredibly prophetic.”
The San Francisco-based company plans to roll out Voice Mode soon to paid subscribers of ChatGPT Plus. OpenAI said the new model can respond to users’ audio prompts “in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in a conversation”.
“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,” OpenAI, a Microsoft-backed company, has posted on X.
Johansson has become the latest high-profile person to take on OpenAI for using creative work without permission.
The company is facing several lawsuits filed by authors and actors — including the Authors Guild of America — for copyright violations.
Scrutiny over the $86 billion start-up continues to intensify as the company tries to challenge Big Tech companies such as Google and Microsoft in developing tools capable of generating responses trained on audio, text and images.
The issue demonstrates the relationship that exists between some tech companies and creatives.
In 2023, Irish actor Remi Michelle Clarke had a contract with Microsoft to voice Bing in Ireland but later discovered a synthetic version of her voice being marketed on another company’s AI voice website. She soon discovered that the contract allowed third parties to use her recordings.
Johansson said “appropriate legislation” is required.
“In a time when we are all grappling with deepfakes and the protection of our own likeness, our own work, our own identities, I believe these are questions that deserve absolute clarity,” Johansson said.
In September, authors Remie Michelle Clarke and John Grisham sued OpenAI for “systematic theft on a mass scale” by using their copyrighted works without permission.
In December, The New York Times became the first major media organisation to sue OpenAI and Microsoft over copyright issues related to its written works.
The lawsuit argues that millions of articles by The Times have been used to train chatbots that now compete with the publication as a reliable source of information.