OpenAI Pulls ChatGPT Voice Over Scarlett Johansson Similarity

OpenAI has pulled one of ChatGPT’s voices after users noted an uncanny resemblance to Scarlett Johansson, just a week after it opened up access to the feature.

Voice Mode enables ChatGPT users to prompt the chatbot using their voice, with the bot generating responses through audio.

Each of ChatGPT’s voices was sampled from actors. However, social media users noted that one of those voices, Sky, sounded like Scarlett Johansson.

One week after the company opened up Sky and the other ChatGPT voices to free users, OpenAI shut it down due to similarity claims.

“We believe that AI voices should not deliberately mimic a celebrity's distinctive voice,” OpenAI said. “Sky’s voice is not an imitation of Scarlett Johansson but belongs to a different professional actress using her own natural speaking voice.”

Johansson voiced Samantha in “Her,” a 2013 movie where Joaquin Phoenix's character falls in love with a chatbot she voices.

In a statement, the actor said OpenAI CEO Sam Altman reached out to her hoping she’d provide the voice for ChatGPT but declined the offer “for personal reasons.”

“Nine months later, my friends, family and the general public all noted how much the newest system named ‘Sky’ sounded like me,” Johansson said. “When I heard the released demo, I was shocked, angered and in disbelief that Altman would pursue a voice that sounded so eerily similar to mine that my closest friend and news outlets could not tell the difference.”

Related:OpenAI Unveils New Model, Widens Access to ChatGPT Tools

Johansson claimed Altman contacted her days before the Spring Update event last week in the hopes she’d reconsider. The actor alleged that before the parties had a chance to speak, OpenAI made Sky public.

Adding to her concerns was a tweet on the day OpenAI showed off Sky and other voices in which Altman referenced the 2013 movie.

OpenAI has since removed Sky, a decision Johansson said they made "reluctantly" after she brought in a team of lawyers.

Johansson has already filed a lawsuit against another AI company for using her likeness. Last November, she sued Lisa AI, an image generation app, for using her likeness in marketing materials without permission.

The “Black Widow” star called for “appropriate legislation” to help protect individuals' rights from being misappropriated by generative AI.

“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 absolutely clarity,” the actor said.

In the wake of the incident, OpenAI said it works with voice actors closely to ensure it takes the right steps for its voice products.

Related:Google, OpenAI New AI Agents Shift Focus to Consumers

“Each actor receives compensation above top-of-market rates and this will continue for as long as their voices are used in our products,” the Microsoft-backed company said.

ChatGPT’s voices were sampled from voice actors the company partnered with to create them, though OpenAI refuses to name them to “protect their privacy.”

OpenAI said it received more than 400 submissions from both voice and screen actors. Each was given a script of ChatGPT responses with samples ranging from answering questions about mindfulness to engaging in conversations about a user's day.

The company said it spoke at length with each of the five actors behind the voices about the risks involved and an internal OpenAI reviewed the voices from a “product and research perspective.”

“This entire process involved extensive coordination with the actors and the casting team, taking place over five months,” OpenAI said. “We are continuing to collaborate with the actors, who have contributed additional work for audio research and new voice capabilities in GPT-4o.”

Beyond the issues with Sky, OpenAI said it plans to introduce new voices in the future to “better match the diverse interests and preferences of users.”

Voice and audio generation is becoming an increased focus for OpenAI. The company unveiled Voice Engine, a text-to-audio application in April, but stopped short of releasing it to the public over misuse fears.