OpenAI has started to roll out an advanced voice mode for ChatGPT to a limited number of ChatGPT Plus subscribers. This new feature initially showcased at the GPT-4o launch event in May, faced criticism for its uncanny resemblance to Scarlett Johansson’s voice and was subsequently delayed due to safety concerns.
At the event, the new voice mode demonstrated significant improvements over the current ChatGPT voice feature. OpenAI employees were able to interact with the chatbot in real-time, interrupting it and asking it to tell stories in different styles, with the chatbot adapting seamlessly. This advanced mode was set to release in alpha in late June but was postponed by a month to ensure it met OpenAI’s high standards for safety and performance.
During the initial demonstration, one of the primary criticisms was the voice’s resemblance to Scarlett Johansson, who played an AI personality in the movie Her. Although the voice had been included in ChatGPT before the spring demo, OpenAI removed it after Johansson sent letters inquiring about its creation. Christianson clarified that the new mode will only use four preset voices made with voice actors. “We’ve made it so that ChatGPT cannot impersonate other people’s voices, both individuals and public figures, and will block outputs that differ from one of these preset voices,” she added.
OpenAI spokesperson Taya Christianson mentioned that the company tested the voice model’s capabilities with over 100 external red teamers to find potential weaknesses. Given the recent scrutiny regarding OpenAI’s safety policies, this delay was a prudent decision. Additionally, new filters have been added to recognize and block requests to generate music or other copyrighted audio.
OpenAI plans to bring the new mode to all ChatGPT Plus users in the fall, according to Christianson.
