You are currently viewing OpenAI Halts Scarlett Johansson-Like Voice After Complaint

OpenAI Halts Scarlett Johansson-Like Voice After Complaint

Rate this post

Scarlett Johansson has criticised OpenAI for using a voice in its new ChatGPT product that is “designed around” her own.

According to Johansson, OpenAI approached her nine months ago and asked if she would be willing to provide her voice for an AI system, but she declined due to personal reasons. She said she was “shocked” and “disturbed” when she heard the voice because it is so similar to hers that even her friends can’t tell the difference between them.

In response to widespread feedback that described the voice as flirtatious and feminine, OpenAI took steps on Monday to remove it from ChatGPT. The company had showcased this voice – called “Sky” – during a recent event demonstrating the new ChatGPT-4o model, and numerous media outlets compared the sound of Sky’s voice to Johansson’s character from the movie “Her”, which premiered in 2013.

Even OpenAI CEO Sam Altman appeared to acknowledge this connection by tweeting the word “her” after the presentation; one week later, OpenAI stated that Sky actually wasn’t based on any specific person or character. The company published a blog post explaining that Sky’s voice belongs to another professional actress who uses her natural voice; they cannot disclose names of their actors for privacy reasons.

Johansson said Altman initially told her that he thought having her voice might help bridge tech companies with creatives and make people more comfortable with AI changes. He asked if she would reconsider two days before they released Sky, but she still declined; then her lawyers contacted OpenAI about getting it taken down.

Many individuals pointed out how much Sky sounds like Johansson’s character in “Her,” including Colin Jost during “Saturday Night Live.” Some have criticised its tone as excessively fawning or gendered.

OpenAI said they chose ChatGPT’s voice based on criteria such as being “timeless” and “approachable and trustworthy.” They reviewed hundreds of voice samples, selecting five different voices for ChatGPT in September. The chosen actors came to a studio in San Francisco where they recorded sessions that were used to train models on their voices.

Shortly after the Sky voice was removed, several members from OpenAI’s top safety team quit – including researcher Jan Leike who said he felt like the company is too focused on “shiny products” over safety. Altman and co-founder Greg Brockman defended OpenAI, saying they wouldn’t release anything if there were concerns about its safety.

In their blog post about creating ChatGPT voices, OpenAI highlighted collaborations with entertainment professionals and how they compensate voice actors. AI companies like OpenAI have faced pushback from entertainers and media corporations regarding copyright issues as well as worries over potential job displacement caused by AI systems; major entertainment unions such as Sag-Aftra have gone on strike over these very matters.

Source: theguardian