ChatGPT is rolling out Live Dictation, which converts microphone input into visible text while the user is still speaking. Tapping or clicking the audio waveform reveals the transcript, and the resulting prompt can be edited before it is sent to the model.

OpenAI is also removing the previous 10-minute recording limit for prompt-box dictation and says transcription reliability has improved. The feature is becoming available globally on free and paid plans across the web, desktop applications, iOS and Android.

Live Dictation is distinct from ChatGPT’s conversational voice mode, where the model responds aloud in a two-way exchange. It also differs from Record mode, a beta feature intended for voice notes and meetings that remains limited to Plus, Pro, Business, Enterprise and Education workspaces.

The update shortens the wait after a long spoken prompt and makes errors visible before submission. It does not remove the need to review names, numbers or specialist terms, and users recording other people still need to follow consent and workplace rules. For simple note capture, the ordinary prompt box can also be asked to turn a dictated message into a transcript or bullet list.