Voice
Dictate messages with your microphone, and listen to replies read aloud in the voice you choose.
Choose voice models
In Settings → Speech & Voice, choose the speech-to-text model for dictation, the text-to-speech model for Read aloud, and a voice if the model offers named voices. These are separate from the model that writes replies. Speech counts toward your plan, and its privacy options are on the same page. Speech doesn’t offer zero data retention, unlike some chat models. See Privacy.
Dictate a message
Before you begin, allow microphone access in your browser or on your iPhone.
- Click Start voice typing (the microphone beside the composer).
- Speak. A timer shows Stop to transcribe · tap composer or Esc to cancel.
- Click Stop voice recording. The transcribed text is added to the end of your draft.
- Review the draft, then send it.
Recording stops after 30 seconds. Cancel discards the recording and keeps your draft.
Bug reports have a separate dictation button, with a one-minute limit and a lower-cost transcription model. See Bug reports.
Listen to a reply
On a finished reply, click Read aloud to generate the audio and play it. Use the same button to replay saved audio. In a regular chat, the audio is also saved to Files → Generated → Audio. In an incognito chat, the audio plays but isn’t saved to Files.