OpenAI introduced ChatGPT Voice on July 23, a voice mode for the desktop version of ChatGPT that allows you to control your computer and give instructions to multiple agents simultaneously in ChatGPT Work or Codex by voice. The feature is already rolling out globally on macOS and Windows for Plus, Pro, Business, Edu, and Enterprise plans.
OpenAI introduced ChatGPT Voice on July 23, a voice mode for the desktop version of ChatGPT that allows you to control your computer and give instructions to multiple agents simultaneously in ChatGPT Work or Codex by voice. The feature is already rolling out globally on macOS and Windows for Plus, Pro, Business, Edu, and Enterprise plans.
ChatGPT Voice is powered by GPT-Live technology, which OpenAI first released for mobile apps in early July. It allows you to listen and speak at the same time, naturally inserting short confirmations like «I understand» without interrupting the user, and leaves the heavy calculations to background models for reasoning.
The architecture separates the voice layer and the execution engines: while a conversation is ongoing, individual agents in Codex or ChatGPT Work can perform tasks asynchronously in the background. This allows, for example, to launch multiple work streams from a single conversation and continue talking while the agents complete their tasks.
Voice mode is available in the ChatGPT desktop app for Plus, Pro, Business, Edu, and Enterprise plans; the latter two will have a two-week early access period before the feature becomes enabled by default. Voice mode can also be used in Codex from the iOS app via a paired connection to a computer in the Remote feature — Android support is promised soon.
According to OpenAI, the launch is a step towards the company’s broader vision: to talk about what needs to be done and get multiple tasks moving in the background «at the speed of thought.» According to Bloomberg, GPT-Live technology could also form the basis of the company’s upcoming hardware device — a home speaker. The Voice mode in Codex is explicitly described by Fortune and Yahoo as a move aimed at developers — one of the most profitable categories of AI users.
The feature works across three desktop app modes — Chat, Work, and Codex — so you can start a new task, check the progress of existing tasks, or adjust your work direction without having to switch back to text input. In a demo video, OpenAI showed two of its own employees — Codex development experience engineer Jason Liu and Codex technical specialist Guinness Chen — speaking simultaneously into the same ChatGPT session in the desktop app, each giving separate instructions.
On macOS, ChatGPT Voice additionally supports the «Appshots» feature, which allows Voice Mode to analyze the active window on the user’s screen, along with local files, codebase structure, and active plugins. Voice Mode also works with OpenAI’s Computer Use feature.
