OpenAI has announced that it has begun rolling out an alpha version of its new Advanced Voice Mode to a small group of ChatGPT Plus users. The new feature implements more natural and real-time conversations, with the ability to interrupt at any time and to sense and respond to the user’s emotions.
Over the next few weeks, OpenAI will gradually expand access to more Plus users, with the goal is to make this feature available to all Plus users this fallThose who participate in the alpha will receive a notice in the ChatGPT app and an email with instructions on how to use it. However, video and screen sharing will not be included in the alpha, but will be launched at a later date.
OpenAI focused on the quality and security of GPT-4o voice conversations to ensure an optimal experience for ChatGPT users. The model’s ability to support millions of cSimultaneous real-time voice conversations have been strengthened, maintaining low latency and high quality.
In June, OpenAI announced improvements to the model’s ability to detect and reject inappropriate content. Over the past two months, the model’s voice capabilities have been tested with more than 100 external red teamers, collectively speaking 45 different languages and representing 29 regions. Based on feedback from 100 external red teamers, collectively speaking 45 different languages and representing 29 regions, and its own internal testing, OpenAI has implemented several safeguards.
Advanced Voice Mode will use four different preset voices, created in collaboration with voice actors, to ensure that ChatGPT cannot impersonate other people, including public figures, by blocking output that differs from these preset voices. Additionally, new filters have been added to recognize and block requests to generate copyrighted music or other audio content. The same security techniques applied to text mode have been extended to voice conversations, ensuring that GPT-4o complies with applicable laws and does not generate malicious content.
