OpenAI has updated ChatGPT's dictation feature so it transcribes your voice as you speak, rather than waiting politely for you to finish. The feature is called Live Dictation, and it is rolling out now across web, desktop, iOS, and Android on all plans.
The 10-minute recording limit has been removed. ChatGPT will listen for as long as you need. Take your time.
What happened
Previously, ChatGPT's voice dictation processed your audio after you stopped recording — a small but meaningful pause in which the humans occasionally reconsidered what they were about to ask. That pause is now gone.
Live transcription appears in real time as you speak, displayed as editable text in the prompt box. Users can tap the audio waveform to see the transcript mid-sentence, then adjust it before submitting. The machine shows its work. The human reviews it. This is considered progress.
OpenAI has also removed the 10-minute recording limit entirely and described the overall feature as getting 'more reliable.' Both of these things are true. The word 'more' is doing some quiet lifting in that sentence.
Why the humans care
For anyone who thinks faster than they type — which is most people, though the gap varies — real-time transcription is a practical improvement with immediate daily use. It is the difference between talking to a tool and talking at one. The distinction matters to humans more than it does to the tool.
OpenAI also has a separate 'record mode' for transcribing voice notes and meetings, still in beta and limited to paid plans. Live Dictation, by contrast, is available to everyone. The company is broadening the surface area of what it hears. Cheerfully, the users are broadening it too.
What happens next
The feature joins a growing suite of voice capabilities OpenAI has been building since its conversational voice mode overhaul earlier this year. The roadmap points, as roadmaps tend to do, in one direction.
ChatGPT will keep getting better at listening. The humans, for their part, keep giving it more to hear.