Meeting transcription software: from the call to text you can use

Meeting transcription software turns what people say on a call, in a meeting room or on a speakerphone into text you can search, quote and send on. Dalm is a Mac app that records the meeting on your computer; nothing joins the call. While people talk, it can show a draft of what has been said. When you stop, the speech is recognised on your Mac and split into lines, each with its voice and its time. No audio is sent anywhere; the transcript is synced to our servers. There is a Free plan.

Where meeting transcription is used: calls, rooms, clients, interviews

A call. Whatever app the call is in, Dalm records two streams on your Mac: your microphone and the sound of the other apps, which is where the other people’s voices come from. That includes a Webex call. Your lines are yours from the start; only the other voices have to be told apart. Nothing is added to the meeting. What this looks like on a Zoom call in particular has a page of its own.

A meeting room. There is no call sound to take. The app listens through the microphone, separates the voices from that one recording and asks you to point out your own — “That’s me”.

A phone call on speaker. It is recorded the way a room is: the Mac’s microphone hears both ends.

A meeting with a client. Nothing shows up among the participants, so the other side will not see a recorder — tell them you are recording. If you planned the meeting in Dalm and invited people, their names are offered when you name the voices, under “From the invited”.

An interview. On a call there are usually two voices: yours, known from your microphone, and the candidate’s. Afterwards you can quote an answer word for word instead of reconstructing it from notes.

How the recording itself works, and which tools leave you a video or a sound file instead, is on the page comparing meeting recorders.

Four ways to get a meeting into text

	When you see text	Who is in the call	Where the audio goes	What you keep
A bot that joins the call	During the call or after it, depending on the product	The bot, listed among the participants	The call is heard on its maker’s servers	What the product keeps, on its maker’s servers
A recording uploaded to a transcription service afterwards	Once the upload is processed	Whatever recorded the call	The file is uploaded to the vendor’s servers	Your recording, and the text in the service
The meeting platform’s live captions	During the call, as captions	No one extra: the platform makes them	The platform already carries the call’s sound	Depends on the platform, its settings and your plan
A Mac app that records and recognises speech on the computer — Dalm’s kind	In Dalm: a draft during the call, the transcript once you stop	No one: the app records on your computer	The speech is recognised on the Mac. In Dalm no audio is sent anywhere, and the transcript is synced to our servers	A transcript split by voice, with times, and a summary

None of the four is right for every meeting. A bot needs nothing on your computer, but it has to be let into the call and has no call to join across a table. Uploading afterwards works with any recording, if you are fine with the file leaving your computer. Captions help people follow along; what stays afterwards is up to the platform. If you need a video or the sound itself, pick a tool that keeps a file — Dalm keeps text.

Apps that recognise speech on the computer are a kind, not a single product, and they differ in what happens to the text next — that is laid out step by step. The kinds of AI note taker are sorted by where the speech becomes text, and named tools, with where each does better than Dalm, are in transcription software compared by name, including the best meeting transcription software for different jobs.

During the call: a draft, not yet the transcript

While a meeting is being recorded, the recording screen can show a column called “What’s been said”, labelled “draft · will be refined after the meeting”. “The first lines appear after a minute or so of speech.” The draft has no names yet — “for now it’s “you” and “them”” — and if the Mac falls behind, it switches off and says so: “The draft is off — we can’t keep up”.

The draft is a preview for following along. It is not saved, and it is not the transcript. Two things you do during the call do last: Add a mark (⌥M) flags a moment to come back to, and the notes you type shape the summary, which is built around what you wrote.

After you stop: the transcript, voice by voice

When you stop, the recording is processed on your Mac: echo and noise are cleaned out, the speech is recognised and split by voice, and every line gets the time it was said. The finished transcript is synced to our servers automatically. The audio is deleted once the text exists.

Naming is optional. Until you answer, a voice shows a guess from the language model or a number — “Speaker 1”, “Speaker 2” — and in the naming step a guess may come marked “Suggested by Dalm — check before you agree.” In the transcript and the summary a guess looks like any other name until you confirm or change it. The summary is written without waiting for names, and a voice is not recognised again in your next meeting. How voices get told apart and named has a page of its own, and so does what the time on each line is for.

Before you use it: names and terms

In a meeting transcript, the words most likely to come out wrong are the ones that matter: a client’s name, a product, a colleague who joined last week. Any line can be edited, and “Show original” brings back what the recognition wrote. Correct a word and the app offers to put it in the dictionary; a term that is already there is swapped in as the meeting is transcribed — “The error is fixed right away, during transcription.” So the names of the people and products in tomorrow’s meeting are worth adding today.

The summary is written once the transcript reaches our servers, and it can be edited line by line; what you change is marked as edited by you. You do not pick the language of a meeting — it is detected — and how the dictionary and the choice of languages work is on the speech to text page; how the languages you speak are covered has a page of its own.

What you can do with the transcript

None of these files has sound in it.

Once the transcript is synced, a third-party language model writes a summary of the meeting from the transcript, your notes and the names you gave the voices; the audio is not part of what the model gets.

Every quote shows you where it was said — open it and the transcript opens at that place. What you land on is a line of text, not a sound.

Two jobs that start from the transcript have pages of their own: minutes drafted from a recording and questions to a recorded meeting. If the meeting is already a file — a platform recording, a voice memo, a video — you can bring it in instead, up to four hours long, on the Free plan: turning a recording into text.

What Dalm does not do

• It does not keep the recording. There is no player and no audio export, so there is nothing to replay; the transcript is what you keep.
• It does not give you a finished transcript mid-call. What you see while people talk is a draft: no names, it can switch off, and it is not saved.
• It does not pause. A recording runs until you stop it, and one you cancel is discarded straight away.
• It does not count very short recordings. With fewer than 100 words, a recording does not count as a meeting.
• It does not recognise a voice from one meeting to the next. Each meeting starts with new voices.
• It does not tell the other people it is recording. It is not in the call, so that is up to you. Before your first recording the app shows a note on recording and consent and suggests saying out loud that you are recording.
• It gives no accuracy figure. We have not published a measurement of English speech with its method, and a percentage without one would say little about your meetings.
• It is not on Windows, and it needs an account. Dalm is a Mac app, and while it is not public, the way in is the waitlist.

Free, Pro or Team: the first meeting you transcribe

Recording, the transcript and the summary come with the Free plan, and Free is a plan rather than a trial: it has no end date. Dalm needs a Mac with Apple Silicon · macOS 15.0 or later and an account. What the paid plans add, and what they cost, is on the pricing page. For a group that transcribes into one shared space — one dictionary, shared templates and tags — see meeting transcription for a team.