OpenAI brings ChatGPT Voice to its desktop app for hands-free agent control

OpenAI has added its ChatGPT Voice mode to the ChatGPT desktop app on macOS and Windows, letting users control their computer and direct multiple AI agents using only their voice. The rollout began July 23 and is going out globally to paying subscribers.

The feature is aimed at hands-free, multi-agent work. Running on OpenAI’s GPT-Live voice model, ChatGPT Voice can listen, speak and coordinate tasks at the same time, so a user can start several work streams from one conversation and keep talking while agents finish jobs in the background.

On macOS, the app adds a screen-context feature OpenAI calls “Appshots,” which lets ChatGPT Voice see what is on the user’s screen. The tool works with ChatGPT Work, the product OpenAI created on July 9 when it merged its Codex coding agent into ChatGPT.

The launch pushes voice from a novelty toward an interface for orchestrating software agents, a direction much of the industry has been chasing. Directing several agents by speech, rather than typing prompts one at a time, is the kind of workflow OpenAI is betting professionals will adopt.

OpenAI said usage of the combined product reached 10 million users, up from 5 million weekly Codex users before the merger. That figure is the company’s own and has not been independently verified, and OpenAI has not detailed how it counts active users.

Whether voice-driven agent control proves reliable enough for daily work, rather than demos, will decide whether it reaches free-tier users or stays a paid feature.