Live captions vs transcription in Teams
On screen they are the same thing: your colleagues' words, appearing as they say them. Underneath they are two different features with different permissions, different consequences and — the part that catches everybody — different answers to "is any of this still here tomorrow?"
The short version
| Live captions | Live transcription | |
|---|---|---|
| Who can turn it on | Any participant, for themselves | Somebody in the organiser's organisation, if policy allows |
| As a guest | Yes | No |
| Who else sees it | Only you | Everyone — it is shown in the meeting |
| Is anyone told | No | Yes, everyone is notified |
| Kept afterwards | No — gone when the call ends | Yes, with the meeting and in the chat |
| Whose it is | Nobody's; it never existed as a file | The organiser's organisation |
| Licence | None | Included in business plans, subject to policy |
Live captions
Captions are an accessibility feature, and every rule about them follows from that. They are for the person who needs to read along, so anyone can switch them on, in anybody's meeting, without permission and without anybody else's screen changing. Turn them on with More (…) → Language and speech → Turn on live captions.
Because they are yours alone, nobody is notified. And because they are not a record, nothing survives the call — close the window and every line is gone. That is the trade: unrestricted, and temporary. The full guide to captions covers changing the language and what to do when the option is missing.
Live transcription
Transcription is a record, and every rule about it follows from that. It produces a file that belongs to the organiser's organisation, so Teams is careful about who may start one: the account doing it has to be in the organiser's tenant, on a licence that includes it, with the tenant's meeting policy and the organiser's own meeting options both permitting it.
Start it with More (…) → Record and transcribe → Start transcription. Everyone in the call is told, immediately and visibly. Afterwards it sits with the meeting and in the chat, and can be downloaded as .docx or .vtt.
Neither of these is Copilot. Copilot summarises a transcript that transcription already made. It writes nothing down itself — see transcribing without Copilot.
Which one you can actually have
For most people the choice is made for them by whose meeting it is.
- Your own organisation's meeting, transcription enabled. Use transcription. It is free, built in, kept for you, and everyone knows it is happening.
- A client's, supplier's or interviewer's meeting. Transcription is not available to you at all, and no purchase changes that. Captions are — and they are all you get.
- Your own organisation, but an administrator has turned transcription off. Ask for it to be enabled; it is a policy setting rather than a cost. Captions work in the meantime.
Which leaves a lot of people watching a complete, correctly attributed transcript scroll past on their own screen, knowing it will be deleted in forty minutes.
Keeping what the captions said
Whispifys is a Chrome extension that saves them. It reads the captions Teams is already showing you and writes each line down as it settles — Teams' own recognition, Teams' own speaker names, timestamped, then searchable across every call you have ever had.
- It needs nothing granted to you — no licence, no admin change, no organiser's permission.
- No bot joins the meeting, so there is no extra participant and nothing to admit.
- The record is yours, in one place, across every organisation whose calls you sit in.
It does not replace transcription where you can have it. It covers the meetings where you cannot.
Captions are not announced — which makes telling people your job. Nobody knows a record is being kept unless you say so, and in many places everyone's consent is required before a conversation is recorded in any form. Our terms put that duty on the person doing the recording, because that is where it sits in law.