ChatGPT Voice Comes to the Desktop App — Control Work and Codex Agents Hands-Free
OpenAI has announced that it is bringing a new “ChatGPT Voice” mode to the ChatGPT desktop app on macOS and Windows. Built on top of the GPT-Live voice model announced on July 8, it lets users launch and direct multiple agents running in ChatGPT Work or Codex using nothing but their voice.
Details
- What it does: Users can control their computer and direct multiple agents running in ChatGPT Work or Codex by voice alone. Because it’s built on GPT-Live, the app can speak, listen, and coordinate work simultaneously
- Context-dependent behavior: The same voice layer behaves differently depending on whether you’re in Chat, ChatGPT Work, or Codex, exposing different tools and permissions in each — in Codex, that means launching and steering coding agents
- Mac-only “Appshots”: On macOS, ChatGPT Voice can be paired with Appshots, which lets ChatGPT reference the content of the active window (including alt-text) for better context
- Rollout: Shipping as ChatGPT desktop app version 26.715, rolling out globally on macOS and Windows simultaneously
- Availability: Available to Plus, Pro, Business, Edu, and Enterprise subscribers; not available on the free plan
- Mobile access: ChatGPT Voice for Codex is also usable from the iOS app via paired remote access, with Android support coming later
How to try it
- Update the ChatGPT desktop app (macOS/Windows) to the latest version
- Requires a Plus tier subscription or above
- Availability may vary depending on plan, region, and workspace settings