Can you transcribe a meeting live, in real time?

Yes. A meeting can be transcribed live, with words appearing as people speak, and some of the options cost nothing. Live transcription is a different job from transcribing a finished recording. The text on screen is a draft that the tool can revise as more audio arrives. Choose the tool by the platform you use, by whether you need a saved record and by how much delay you can accept.

Free live captions are built into Mac and Windows, with limits

Two free options come with the operating system. Apple's Live Captions on Mac processes audio on the device ¹, and it is available only on Mac computers with Apple silicon ². Apple also says its accuracy may vary and that it should not be relied on in high-risk or emergency situations ³.

Windows 11 offers a similar feature. Microsoft says all processing of audio and caption generation happens on the device ⁴. The captions are not kept, because Microsoft says generated captions are not stored anywhere on the device or in the cloud ⁵. That makes live captions a reading aid, not a record. If you need the words saved, use a recorder or a tool that writes the transcript to a file.

Meeting platforms provide live transcripts for their own calls

If everyone is on one service, its live transcript is usually the simplest route. Live transcription in Microsoft Teams is available only on the Teams desktop app ⁶. For Google Meet and Zoom, check the admin and plan settings in our guide to Zoom, Teams and Google Meet transcription. The free meeting transcription question covers plan limits too.

Streaming services are built for live audio, but they reconnect on a timer

Developers can call streaming speech services that return results while the speaker is still talking, and the interim results can still change before the final one arrives. Streaming has a limit on length. Google Cloud's quota table gives a streaming recognition request an audio length of about 5 minutes ⁷, so a long meeting needs code that reconnects. Use a streaming API only if a developer can build that part.

Whisper needs a workaround for live audio, and the output lags

Open-source Whisper can run live, but only with a workaround that cuts the sound into short pieces. The whisper.cpp stream example does this for you: it samples the microphone every half a second and runs the transcription continuously ⁸. It is a free option for a technical user, and it will need testing before a real meeting. Our question on whether Whisper can transcribe in real time explains the lag and the revisions in more detail.

Live text changes, so keep the final transcript

A live caption is a draft. In Artificial Analysis's 2026 streaming benchmark, OpenAI's hosted GPT Realtime Whisper model improved from a 7.5 percent word error rate on the first partial transcript to 5.1 percent on the final transcript ⁹. Word error rate is the share of words a system gets wrong, where lower is better, as the word error rate entry explains. Those figures come from one hosted model in one benchmark, so use them to see the pattern, not to predict your meeting. The words you read live may change, so read the final transcript for anything that matters.

What to do

1. Decide whether the live text is for reading along or for a record. If it is a record, keep the audio.
2. Use the platform's live transcript if everyone is on that platform.
3. Use Apple's or Windows' built-in captions for a free reading aid, and check that your language is supported first.
4. Test on a short call, watch for lag, and see whether the captions keep up with overlapping speakers.
5. Announce the transcription before the meeting starts, and check the consent steps in the meeting transcription guide.
6. After the meeting, read the final transcript against the audio at each decision.

For more on free options, read free meeting transcription and transcribing a meeting without a bot.