ChatGPT Debuts a New Voice — But It Comes With a Catch In May, OpenAI demoed the new voice mode but some felt it sounded too close to Scarlett Johansson's — including Johansson, who hired legal counsel.
By Sherin Shibu Edited by Melissa Malamut
Key Takeaways
- As of Tuesday, a small group of paying customers can access OpenAI's new voice mode.
- Voice mode took front and center stage in OpenAI's GPT-4o May demo for its responsiveness and emotiveness.
- OpenAI had to work through possible safety issues before rolling out the feature.
After a one-month delay, OpenAI is giving ChatGPT a human voice.
ChatGPT, which takes a written prompt and churns out an answer based on what it knows from its training data, was previously limited to typed answers. As of Tuesday, a limited group of paying ChatGPT Plus subscribers now have access to another dimension of the AI chatbot: They can access four pre-loaded voices to talk to ChatGPT and get answers in real time.
Related: OpenAI's Launches New AI Chatbot, GPT-4o, Which Sounds Almost Like a Friend Would
With the voice mode, paying users can talk to ChatGPT, interrupt its answers, and have more natural, human-like conversations. Each voice "senses and responds to your emotions," according to a Tuesday post on X from OpenAI.
We're starting to roll out advanced Voice Mode to a small group of ChatGPT Plus users. Advanced Voice Mode offers more natural, real-time conversations, allows you to interrupt anytime, and senses and responds to your emotions. pic.twitter.com/64O94EhhXK
— OpenAI (@OpenAI) July 30, 2024
In May, OpenAI demoed the new voice mode — to varying reactions. OpenAI CEO Sam Altman said the advanced voice assistant felt "like AI from the movies" but movie star Scarlett Johansson was shocked by how similar one of the voices sounded to hers, and hired legal counsel.
OpenAI also had several high-profile resignations over safety concerns following the demo, with its chief scientist leaving to start his own safe AI company.
Though voice mode was supposed to arrive in late June, OpenAI delayed its launch and said it needed more time to scale the technology safely to millions of users.
In the X release thread, OpenAI said it tested GPT-4o across 45 languages and only allows the model to speak in preset voices, "to protect people's privacy."
Sam Altman, OpenAI CEO. Photographer: Chona Kasinger/Bloomberg via Getty Images
Voice mode will also block copyrighted and violent requests from users.
"We plan to share a detailed report on GPT-4o's capabilities, limitations, and safety evaluations in early August," OpenAI stated.
ChatGPT has more than 180 million users as of July 2024, with around 3.9 million paying subscribers in the U.S.