OpenAI announced an update to the ChatGPT desktop application (now integrated with Codex), introducing a ChatGPT Voice mode powered by the GPT-Live model. Users can initiate, check, or adjust tasks through voice in three areas: chatting, working, and programming, enabling multi-threaded work collaboration.

Full-duplex voice support, talk and work at the same time

In terms of development, OpenAI first introduced voice recognition features for ChatGPT in 2024, and has since continued to invest in voice technology. The newly released GPT-Live model achieves full-duplex voice, allowing simultaneous listening and speaking, making AI conversations feel more like real human interactions. OpenAI stated in their announcement: "We are moving towards our AI vision, and in the future, users will only need to speak their requirements, and ChatGPT will quickly advance multiple tasks based on your thoughts."

ChatGPT Voice allows users to start multiple workflows from a single conversation and continue talking while intelligent agents complete tasks in the background. For example, when planning a business trip, users can ask ChatGPT to check the calendar for conflicts, sort through the inbox to see if there are any changes to flights, and prepare meeting notes—all of which are completed while "the user is making coffee or doing other things." This means AI is no longer just a tool that answers one question at a time, but rather something that can truly take over your multi-task workflow, moving voice interaction from simple conversation to efficient agent collaboration.