OpenAI has decided to discontinue the use of a voice from ChatGPT identified as Sky after it was noted by users to bear a resemblance to Scarlett Johansson’s voice. The organization made this decision public on Monday, following the use of this voice during the demonstration of their latest GPT-4o model.
The company stated, “It is our stance that AI-generated voices should not intentionally replicate the unique voice of a celebrity. The voice known as Sky was not crafted to imitate Scarlett Johansson; it is the genuine voice of another professional actress.” OpenAI mentioned in a blog post that they are unable to disclose the identities of their voice actors to safeguard their privacy.
Social media buzzed last week with a video from the demonstration, where many users commented on the similarity of the voice to Johansson’s. The voice received mixed reactions, with some criticizing it for seeming too flirtatious and others comparing it to an idealized male vision.
The seductive tone of the voice has been likened to the 2013 movie “Her,” where Johansson provides the voice for a provocative digital assistant. In the movie, the protagonist, portrayed by Joaquin Phoenix, develops romantic feelings for the digital assistant.
While OpenAI has not made any direct comparison between Sky’s voice and that of Johansson’s, OpenAI CEO Sam Altman did reference the movie “Her” in a tweet after the company’s presentation.
The demo aimed to showcase the chatbot’s enhanced conversational abilities, but went viral after the sultry voice giggled at almost everything an OpenAI employee was saying. At one point, the chatbot told the employee: “Wow, that’s quite the outfit you’ve got on.” At another point, the chatbot said “Stop it, you’re making me blush” after receiving a compliment.
In its blog post, OpenAI says it wants its chatbots’ voices to sound “approachable” and to “inspire trust.” It also wants them to have a “warm, engaging, confidence-inspiring, charismatic voice.”
Going forward, OpenAI says it plans to “introduce additional voices in ChatGPT to better match the diverse interests and preferences of users.”