Farewell to ChatGPT's Imitation Scarlett Johansson Voice

Last week, OpenAI unveiled GPT-4o, featuring a more human-like voice mode. Observers noted that one of ChatGPT's voices resembles Scarlett Johansson, particularly her character in the film Her. OpenAI stated that the similarity between the flirtatious AI voice, named Sky, and Johansson was unintentional. The company announced it is "pausing the use of Sky" while addressing concerns regarding the voice.

In a blog post, OpenAI clarified, "We believe that AI voices should not intentionally mimic a celebrity's unique voice. Sky’s voice is not an imitation of Scarlett Johansson but belongs to a different professional actress using her own natural speaking voice." They also highlighted that due to privacy considerations, they cannot disclose the identities of the voice talent. Furthermore, OpenAI emphasized that each performer is compensated "above top-of-market rates," a policy that will remain as long as their voices are integrated into their products.

Interestingly, shortly after the demonstration of the upgraded Sky voice, CEO Sam Altman tweeted the word "her." However, the company firmly stated that this is "not an imitation."

Scarlett Johansson’s portrayal of Samantha in Her is one of the most recognized representations of a virtual assistant in cinema, predating the conversational AI trend by about a decade. It’s no surprise that her performance has become a template for modern voice assistants. Johansson has previously taken legal action against a developer for using an AI-generated version of her voice and likeness without permission.

As of now, details regarding the specific reasons behind Sky's temporary withdrawal or any intended changes before its reinstatement remain unclear.