OpenAI Begins Launching Her-Inspired Voice Feature for ChatGPT

OpenAI's advanced voice mode for ChatGPT is gradually being rolled out to select ChatGPT Plus subscribers. First showcased at the GPT-4o launch event in May, this feature faced criticism for resembling Scarlett Johansson's voice and was temporarily delayed for safety enhancements.

At the launch, the new voice mode demonstrated improved responsiveness compared to the existing one, allowing OpenAI employees to engage with the chatbot flexibly, prompting it to tell stories in various ways. This capability indicates a significant upgrade in user interaction.

Initially scheduled for an alpha release in late June, the rollout was postponed by a month to ensure it met the company's quality standards. OpenAI cited enhancements in the model's ability to detect and refuse inappropriate content as a reason for the delay. Taya Christianson, an OpenAI spokesperson, shared that the voice model underwent rigorous testing with over 100 external red teamers to identify potential vulnerabilities. The company has also implemented new filters to block requests for generating copyrighted audio.

A major critique during the event centered on the new voice, referred to as “Sky,” which bore a striking resemblance to the voice of Johansson, who portrayed an AI character in the film Her. While this voice had previously been part of ChatGPT, it was removed ahead of the event after Johansson raised questions about its creation. To address concerns, Christianson stated that the updated voice mode will exclusively feature four preset voices created with professional voice actors. She emphasized that ChatGPT is designed not to impersonate the voices of individuals or public figures, blocking any outputs that deviate from these preset voices.

OpenAI aims to roll out the new voice mode to all ChatGPT Plus users this fall.