ChatGPT voice mode supports live conversation in the desktop app. Learn how Voice differs from dictation and when typed chat may be the more practical choice.
What ChatGPT voice mode does
ChatGPT Voice provides live spoken conversation in Chat, Work, and Codex within the ChatGPT desktop app. It supports conversational turn-taking: you can interrupt a response, ask a follow-up question, check progress, or redirect the task while the exchange continues.
Voice can also coordinate supported tasks. The documentation says it can start separate tasks for longer work, check existing tasks, and deliver follow-up instructions. It can report progress, blockers, and results in the voice conversation. A destination task must support Voice and be available on a connected host.
This makes Voice a reasonable option when speaking feels more natural than composing successive messages. As an editorial suggestion, consider it for brainstorming, exploring an idea through follow-ups, or steering ongoing work. These are suggested uses, not guarantees that Voice will improve a particular result.

Voice, dictation, and typed chat serve different workflows
Voice and dictation are documented as distinct features. Voice is for a live conversation, while dictation changes speech into prompt text before the prompt is sent.
The desktop prompting documentation adds an important detail: selecting Dictate in the composer transcribes speech there, where the text can be reviewed and edited before sending. Typed prompts can likewise express a question, instruction, or goal, and follow-up messages can refine the response.
As a practical suggestion, choose Voice when you want an ongoing spoken exchange. Choose dictation when you prefer speaking but want an opportunity to review the resulting prompt text. Consider typing when exact wording, detailed boundaries, source instructions, or a required output format matter enough that you want to compose them deliberately.

How to start and stop Voice
In a new desktop chat, select the Voice icon before sending the first message. The first use requires microphone access and a voice selection; on Mac, you are also asked to review Screen context.
For an existing supported task, select Start voice chat. A new chat or task can instead offer Start new voice chat. Select Stop voice chat when finished. An earlier voice conversation can be reopened and resumed with Start voice chat. A keyboard shortcut can be configured under Settings > Voice > Voice chat hotkey.
Voice is documented for the ChatGPT desktop app on Plus, Pro, Business, Edu, and Enterprise plans. It can also be used through Codex on iOS after the phone is paired with a desktop host. Availability additionally depends on rollout and workspace settings.
Access, troubleshooting, and screen context
Only one voice chat can be active across the desktop app at a time. If a chat will not start, the official troubleshooting guidance says to confirm availability for the plan, rollout, and workspace, check microphone permission, and determine whether another app window already has an active voice chat.
A narrower update step applies when Start voice chat is missing from an existing task: update the desktop app and the Codex host running that task. Because Voice in existing Codex tasks is rolling out, a new voice chat or dictation may remain available where supported.
On macOS, Screen context can take an appshot of the frontmost window. That context may include the window image and accessible text beyond the visible scroll area. The provider advises avoiding windows containing sensitive information. As a practical suggestion, inspect the frontmost window before asking Voice to look at it.
Conclusion
ChatGPT voice mode is best understood as a live conversation interface, not merely speech converted into a message. It supports interruptions, follow-ups, and changes of direction while preserving the conversation’s context.
My practical recommendation is to use Voice for conversational exploration and active steering. Use dictation when you want to speak and then review or edit the transcribed prompt. Prefer typed chat when you want to formulate precise goals, context, output requirements, or boundaries before sending. For consequential work, the prompting guidance recommends reviewing the result before using or sharing it.
Before starting Voice, confirm that it is available in your desktop app and that microphone permission is enabled. If you plan to use Screen context on macOS, first bring forward only material you are comfortable sharing.
Sources
Disclosure
This article was prepared with AI assistance.
Improve this guide
Report a problem or suggest the next question
Submissions enter a private editorial queue. They are reviewed and deduplicated before any public update or new page is created.
