OpenAI has started rolling out GPT-Live, a new voice mode for ChatGPT that allows the AI to listen and speak simultaneously, making conversations feel more natural while also performing web searches in the background.
The new feature replaces the traditional turn-based voice interaction, where users and the AI had to wait for each other to finish speaking before responding. With GPT-Live's full-duplex architecture, ChatGPT can process incoming speech while generating responses, allowing for interruptions and a more fluid conversation similar to a regular phone call.
According to OpenAI, GPT-Live continuously processes audio input while speaking, enabling it to better determine when to respond, pause, interrupt or simply listen.
The company is launching two versions of the new model: GPT-Live-1 and GPT-Live-1 mini.
Background web searches
Another major enhancement is GPT-Live's ability to perform web searches while continuing the conversation.
Previously, ChatGPT's voice mode paused the conversation whenever it searched the internet. The new system allows search agents to work in the background, reducing interruptions and improving response times.
Growing competition in AI voice
OpenAI joins a growing list of AI companies developing more human-like voice assistants.
Startups such as Thinking Machines and Sesame AI have already demonstrated full-duplex voice models capable of listening while speaking and conducting background searches. GPT-Live now brings similar capabilities to ChatGPT's global user base.
Availability
OpenAI said GPT-Live is rolling out gradually to all ChatGPT users, including those on the Free plan.
The company noted that availability will expand over time, meaning some users may receive the feature later than others.
