Will Dalm run on your Mac?

Everything the app needs from your machine, in one place: processor, system, permissions, network and what the first launch downloads.

Dalm runs on Macs with Apple Silicon (M1 or later) and macOS 15.0 or newer. It does not run on Intel Macs, on Windows, or on Android. There is no self-hosted version. The first launch downloads a speech model, so the app needs a network connection and a few minutes of patience before it is ready to use.

What you need

WhatWhat is neededHow we know
ProcessorApple Silicon (M1 or later). Intel Macs are not supported and no Intel build is producedRead in the build settings: the architecture list holds arm64 and nothing else
SystemRequires macOS 15.0 or newerRead in the build settings. Not exercised on 15.0 itself — see below
macOS permissionsMicrophone and Accessibility when you set the app up: Accessibility is what lets dictated text land in the app you are typing in. Permission to record system audio, called “Screen & System Audio Recording”, is asked before your first meeting. Calendar and notifications are optional. There is no separate Input Monitoring requestRead in the app’s permission strings and Info.plist keys, product export of
InternetNeeded to sign in — with Google, or with a code or a link sent to your email — to sync, for summaries and documents, and for the first download of the speech models; dictation, recording and transcription carry on without it. Speech is recognised on your Mac by default, and the text is synced to our servers automatically once you are online, with no setting that turns the sync off. The app does not open without signing inConfirmed in the product code, with the fields named

The processor, system and network lines were read in the product code on , and the network line was confirmed again by the developer nearly two weeks later. The permissions line comes from the product export of .

One of those lines deserves its caveat spelled out. Requires macOS 15.0 or newer is a requirement, not an observation: the build is produced on a later macOS and the measurements were taken on a later one still, so nobody here has started the app on 15.0. We write it as the floor because that is what the code declares, and we say plainly that we have not stood on it.

What happens on first launch

There are two moments of waiting, and the second one catches people out because it does not arrive at install time.

WhenWhat downloadsHow big
First launchSpeech models, European family, twenty-five languages461 MiB
First launchSpeech models, all languages, 99 languages600 MiB
First meeting you processSpeech detection and speaker separationabout 22 MB

The language counts describe what the models cover; we have not verified quality language by language, and the numbers are not a promise about any particular one.

What is not supported

QuestionAnswer
Intel MacNo. The app is built for Apple Silicon only
WindowsNo, and it is not planned
AndroidNo, and it is not planned
Self-hostedNo

Using two Macs

Your account carries the text. Transcripts, notes and the names you gave to voices appear on both machines, because that is the part that is synced.

Voice recognition does not carry over, and the reason is wider than the second machine: voices are identified afresh in every meeting, on either Mac. What the app works out about a voice belongs to that one recording and is not carried into the next one, so a colleague you named last week is a voice to be identified again this week — on the same machine you named them on.

Nothing about the second Mac makes this worse, then, and nothing about the first one makes it better. Names, once you give them, are text: they live in your account and appear on both machines. It is the identifying that starts from nothing each time.

Signing out of a Mac removes everything the app keeps locally on that machine. What is in your account stays there.

The reason is the data boundary rather than a limitation chosen for this page: how it works sets out which pieces of data live where.

Where to go next

  • Download — the file itself and what happens after you install it.
  • How it works — what is computed on your Mac by default and what the server builds from the text.
  • What we do with your recordings — the commitments, with dates.
  • The Mac app — what each part of the Mac does in it.

Questions

Does Dalm work on Intel Macs?
No. The app is built for Apple Silicon only, and no Intel build is planned. On an Intel Mac there is nothing to install, so the answer does not depend on which macOS you are running. If you expect to move to an Apple Silicon machine later, the waitlist is the place to leave an address, and we will write when the beta opens.
Does Dalm work on Windows?
No, and it is not planned. The app leans on frameworks that exist only on macOS, so a Windows version would be a different product rather than a port of this one. The same answer covers Android. There is no self-hosted version either, and none is planned: the server side is not something we hand out to run elsewhere.
What macOS version do I need?
Requires macOS 15.0 or newer. Treat that as a requirement rather than a tested guarantee. The build is produced on a later macOS and the measurements were taken on a later one still, so 15.0 itself has not been exercised by us. If your Mac is below that line, updating macOS is the fix.
How much does it download on first launch?
The European family of models, twenty-five languages, is 461 MiB. The all-languages family, 99 languages, is 600 MiB, the same on every Mac. A further 22 MB for speech detection and speaker separation arrives when you process your first meeting, not at install time.
Does Dalm work without internet?
Partly. Once the speech models are downloaded, dictation, recording a meeting and transcription keep working with no connection, because speech is recognised on your Mac by default. Signing in, syncing, summaries and documents need the network: the app does not open without signing in, and the text goes to our servers automatically once you are back online, with no setting that turns it off. If your very first meeting happens offline, it is kept but transcribed as one voice: the speaker-separation model has not arrived yet. The full list is under what works without a connection, line by line. If you choose server recognition in Settings, the audio is sent to Dalm and our speech recognition providers and deleted once the text is ready.
Can I use Dalm on two Macs?
Yes, and what carries over is the text. Transcripts, notes and the names you gave to voices live in your account and appear on both machines. What does not carry over is voice recognition — though that is not really about the second Mac: voices are identified afresh in every meeting on either machine, so a colleague you named last week is a voice to be identified again this week, on the same Mac you named them on.

Version and changes

Version 2 · Last changed

  • — Added the macOS permissions the app asks for and the ways to sign in.
  • — Page built.