Meeting notes that never leave your Mac.
No bot joins your call. No account. Nothing uploaded. Dictera hears both sides through macOS itself and writes the notes locally - so you can take notes in the meetings the other tools aren't allowed in.
Free 14-day trial · No account needed · macOS 26 or later
How a meeting goes
Start listening when the call begins. The transcript builds live on the left. Press Stop and the notes are written - or press summarize at any point and get them mid-meeting.
Nothing joins your call
The other participants are captured with a Core Audio process tap - the same audio you're already hearing. No invite, no bot in the participant list, no Screen Recording permission, and the call keeps playing normally.
One click mutes you
Your microphone starts on, so what you commit to reaches the action items. Turn Mic off when your own voice shouldn't be part of the notes; the panel shows which state you're in throughout.
Pause and resume safely
Stopping twice won't stack a second, identical set of notes - only new speech since the last set counts.
Cancel costs you nothing
A run you stop halfway keeps nothing and leaves the meeting ready to summarize again. Half a set of notes is worth nothing.
Or let the calendar start it
The meetings people forget to record are the ones that started while they were still finding the window. Connect your calendar - read-only, from the accounts already in Calendar, with no second sign-in - and Dictera offers to record each one as it begins, then files the notes under the meeting they belong to.
A countdown, then it records
About a minute before a meeting, a countdown asks whether to record it. Left unanswered it starts, because a meeting you didn't record is gone and one you didn't want is a click to delete. Set silence to mean "no" instead, or turn the countdown off - globally, or for one series.
Never without showing you
The countdown is the consent surface. If it can't be put on screen, nothing records - and skipping the same meeting twice makes Dictera stop asking about that one for good.
The invite fills in Meeting Setup
Who was invited becomes the participant list action items are attributed against, without retyping it every week. Email addresses are never stored, and one switch keeps those names on local models only.
Filed under the meeting it repeats
A recurring event files each occurrence under one series, and keeps doing so after somebody renames the meeting in Calendar - the label is yours from the first time you see it.
Read-only, and off by default
Dictera never writes an event and never puts a bot in a meeting. Calendar access isn't asked for at first run: nothing is read until you switch it on and pick which calendars count.
Today, at a glance
The session list can show the day's agenda - recorded, still coming, or missed - with a month grid for any other day.
Four things it can write
The same recording, at four levels of fidelity. What separates them is only what each is allowed to throw away - so pick the one that matches why you recorded. Both the Meeting button and Stop produce whichever you chose, and the other three stay one click away on the button's menu.
Notes
The outcome: summary, decisions, action items, open questions. Written automatically when you press Stop.
Full minutes
The whole meeting in order - every point, argument and figure - at roughly a third of the length. Not a summary.
Transcript
The actual words, cleaned of disfluencies and mis-hearings, in the language they were spoken. Never translated, never condensed.
Chapters
The shape of the meeting: a timeline of what was discussed when, with times taken from the transcript rather than estimated.
Notes that match the session
A lecture has no decisions and no owners. Forcing "summary / decisions / action items" onto one manufactures action items out of "read chapter four" - so Dictera reads each recording for what it actually is.
Six shapes, detected automatically
Meeting, lecture or talk, one-on-one, interview, sales call - or pick one yourself. The notes state which shape they used, so a wrong guess is one click from fixed.
Decisions are never quietly dropped
Sections that matter say "None." rather than vanishing, so you can tell "nothing was decided" from "the model forgot".
Reshape in one press
Cut to Action items, reorganize By topic, or turn the notes into a Recap email - a finished deliverable, not a draft. Lectures also offer Key terms and a Study guide.
Not offered where it would lie
Rewrites are deliberately unavailable on the transcript and chapters. A reshaped transcript is no longer a record.
Or edit it yourself
A write-up is a document, not a reply. Fix a misheard name, drop a line nobody agreed to, add what the model missed - without regenerating the nine-tenths that were right. The model's own version is kept so Revert works months later, and the exported file discloses that a human edited it.
Who said what, honestly
Everyone except you arrives on a single audio channel, so the transcript can't tell voices apart. Dictera doesn't pretend otherwise - it works from who you say is in the room, then uses those names only where the meeting itself makes it clear.
Meeting Setup
Name the other participants, say who you are, add any background - or let the calendar invite fill it in. New meetings start from your defaults, so a recurring one is typed once; from the first edit the setup is that session's own, and reopening an old meeting shows the people who were actually in it.
"Unassigned", not a guess
An action item whose owner the transcript never establishes is written as Unassigned. This is a rule in the prompt, not a hope.
Background fixes proper nouns
Project, product and people names the transcriber mangles come out right once they're in the setup.
No voice fingerprinting
Dictera never builds a voiceprint of anyone. Attribution is inference from names spoken aloud, and nothing else.
Long meetings
An hour-long call is longer than most models can read at once. Rather than truncating quietly, Dictera reads it in passes and tells you what it covered.
Read in passes
The panel says which pass it's on, and the menu tells you how many a given output will take before you commit to it.
It says when it fell short
If a meeting outruns what one set of notes can cover, the result is labelled as the most recent stretch rather than presented as the whole thing.
The end is what's kept
When something has to go, it's the oldest speech - because decisions land at the end of a meeting.
It won't run all afternoon
A recording that outlives its meeting, or that has simply been going for hours, gets a one-minute countdown and a Keep recording button. Unanswered, it stops - and writes the meeting up on the way out, so nothing is thrown away either way.
A bigger window covers more
Point Meeting Notes at a model with a larger context window and the same call is read in fewer passes. Connection tests fill the window in automatically.
Afterwards
The meeting doesn't disappear when the notes are written - and neither does the series it belongs to.
Ask it about the meeting
Type a question and Dictera finds the passage that answers it, even in a two-hour call - and says plainly when the transcript doesn't settle it, instead of concluding from silence.
What the whole series still owes
Every open action item across every meeting filed under one series, in one list. It reads back the notes you already have - no model runs, so it can't invent a commitment - and it says so when some meetings couldn't be read.
Save as Markdown
Notes, minutes, transcript and chapters all export to a file - with the date, the participants and the model that wrote them - plus copy buttons for each.
Searchable archive
Sessions are listed, searchable and renameable in place, with a title generated for each, and can be filed or deleted in batches. Or switch history off and nothing is written at all.
Stays on this Mac
Transcripts, notes and any screenshots live in your Application Support folder and are deleted with the session. Nothing is synced anywhere.
What it doesn't do.
Does it label each participant separately?
No. Everyone but you shares one audio channel, so the transcript has two labels: you, and them. Names appear only where the participants are listed - typed by you or read from the calendar invite - and the meeting makes clear who is speaking. There is no voice fingerprinting.
Are the notes written live?
No - only the transcript is live. The notes are written when you press Stop, or when you ask for them. That's deliberate: a set of notes that rewrites itself every thirty seconds isn't a document you can read.
Does it handle a multilingual meeting?
Each session transcribes in one language that you choose. The assistant doesn't auto-detect mid-call, so a meeting where people switch language is transcribed in the one you picked. (Automatic detection is available for dictation - it's a different pipeline.)
Can I play back the audio?
No. Dictera never writes the audio to disk - the only artifacts are the transcript and what you generate from it. Timestamps tell you when something was said; there's no recording to scrub through.
Will it record meetings from my calendar?
It will never join one - there's no bot, and Dictera never writes to your calendar. But if you connect Calendar, it can start a recording on your own Mac: about a minute before a meeting, a countdown asks whether to record it, and if nobody answers, it starts. You can set silence to mean "no" instead, or turn the countdown off, globally or for one series. Nothing starts without the countdown being shown first, and skipping the same meeting twice stops Dictera asking about that one. Calendar is off until you switch it on.
What does Dictera read from my calendar?
The title, the time, whether it repeats, and the names on the invite - read from the accounts already set up in Calendar, so there's no second sign-in and no separate connector. Access is read-only. Email addresses are never stored at all, and one switch keeps attendee names on local models only. You choose which calendars count.
Is it really all on-device?
With the built-in model, yes - transcription is Apple's on-device speech engine and the notes are written by a local model. If you point Meeting Notes at a cloud provider instead, the transcript window and any screenshots go to that provider on each pass. The app tells you before you switch. See Security & data handling for the full picture.
Do I need to tell people I'm taking notes?
Recording laws vary and some places require everyone's consent. Dictera captures the audio your Mac is already playing, but that doesn't change your obligations - check what applies where you and the other participants are. It matters more if you let the calendar countdown start recordings for you: the meeting then opens with a recording running that you didn't press Start on.
For the meetings other tools aren't allowed in.
Included with every plan, unlimited during the 14-day trial.