Voice to text on your Mac, straight into the app you are typing in
Voice to text means speaking instead of typing and getting the words as text. With a web page or a phone app, you speak into their box and copy the words to where they are going. On a Mac, Dalm puts them there directly: you press a key, speak and press it again, and the text lands in the field where your cursor is, in any app with a text field. The speech is recognised on your Mac and by default no audio is sent anywhere; the text of every dictation is synced to our servers automatically. There is a Free plan.
Set up once: permissions, a microphone check, a first try
The first launch is a short setup — in the app’s words, “two permissions, a microphone check, a trial dictation and signing in”. You confirm you are 16 or older and accept the terms. Then come the permissions: the microphone, and Accessibility, which lets the text land in whatever app you are working in. macOS opens System Settings at “Privacy & Security” → “Accessibility” for you, and the app restarts by itself to pick the permission up; a third one, for recording other apps’ sound, is asked for only before your first meeting. You say something and watch the waveform move, answer “What language are your meetings in”, and the speech model downloads — it keeps going if you close the window, and dictation switches on as soon as it is ready. Then you try it: press the key and say “This is my first note with Dalm”. Last, you sign in: the app does not work without an account.
Each time: press, speak, press again
Press the right ⌘ key and speak; press it again, and the words appear in the field where the cursor is — a message, an email, a document, a form in the browser. The app puts them in through Accessibility, and if a field does not accept that, through the clipboard and ⌘V; the clipboard is restored after 300 ms. Punctuation and capital letters come from the recognition model. Pressing another key while you hold — ⌘V, for instance — throws the dictation away, which is how the right ⌘ can still do its usual job. In settings (the Shortcuts tab) the key can be swapped for any single modifier, Caps Lock, fn, or a key with a modifier, and holding the key instead of pressing twice can be switched on there. There is no language to pick: it is detected as you speak, within the range you chose at setup.
Longer text: speak as long as you need
For a paragraph or more, nothing changes: the recording carries on hands-free until you press the key again, and then the text is put in. While it records, you can keep typing. A preview of the words can show in the menu bar as you speak; it is a draft rather than the final text, and on dictations longer than about a minute it switches off. The widget — a narrow pill that never takes the focus — shows a waveform and a timer, by default only while you record; it can be set to Always or Never.
When the words do not land where you expected
- Nothing appeared in the field. The dictation is saved anyway, and the widget says “Couldn’t paste — copy the text with the button below”. Copy it from there.
- A meeting is being recorded. While a meeting records, dictation does not start and the key does nothing; the widget keeps showing the meeting’s timer. Stop the recording, then dictate.
- You want the text kept but not typed. Turn off “Insert text into the active app” in settings: the recording is still saved and shows up in the list.
- The waveform did not move. “Check which input is selected in the system sound settings.” If another app is holding the microphone, you get an error that says so rather than silence.
- You switched microphones mid-sentence, from the built-in one to headphones: the recording carries on, with a short gap at the switch.
- Nothing was recognised. The dictation is saved with the title “No text”.
Fixing a word that keeps coming out wrong
Every dictation lands in the Voice typing list, grouped by day, where you can edit it, copy it and search its full text. When a name or a term keeps coming out wrong, correct it in the list; in a meeting, you can correct any line of a transcript. How the list works in more detail, and whether the dictation built into macOS would do for you, is on the dictation page.
Where your voice goes, and where the text goes
The speech is recognised on your Mac; by default, no audio is sent anywhere, and the audio is deleted once the text is ready. The text is another matter: each dictation is synced to our servers automatically, along with its title and the way you started it, so that it is there on every Mac you sign in to, and the sync has no off switch. A dictation is not sent to a language model — it goes through recognition, and that is all. Nobody else sees your dictations, even in a team: not an admin, not the owner. A dictation you delete is gone at once, without a trash, and settings can delete all of them. What keeps working with no connection is on a page of its own, and what goes where, piece by piece, on privacy.
Voice to text for messages, documents and prompts
The field decides how you use it. A message or an email: press, say it, press again, read it once. A document: press to start, talk through a paragraph, press again, then edit. A request to a language model — in a chat window or an assistant — can run long; press once, think out loud until it is done, and press again. Dalm inserts the text and formats it for the app you’re typing in. Dictating to Claude Code, and how it compares with the built-in /voice, has a page of its own. Describing a whole task to a coding agent this way is on voice coding for developers. The model receives only the text that lands in its field: Dalm itself does not send your dictation to a model. Names of tools, libraries and colleagues are worth a look afterwards: you can correct any line.
Free, Pro or Team: starting with your voice
Dictation comes with the Free plan, and Free is a plan rather than a trial that runs out. Dalm needs a Mac with Apple Silicon · macOS 15.0 or later and an account; while it is not public, the way in is the waitlist. What the paid plans add is on the pricing page. In a team, transcripts, notes, documents and tags are shared while each person’s dictations stay private to them — working as a team. The same app also records calls and meetings in the room: one app for both jobs, or two.
Questions
- What is the best voice to text app for Mac?
- It depends on where the words should end up. If you want them typed straight into the app you are using, you need voice to text software that works system-wide rather than a web page; macOS has dictation built in, and how it compares is set out separately. Dalm is a Mac app that types into any app with a text field; it needs Apple Silicon · macOS 15.0 or later, and while it is not public, the way in is the waitlist. There is no Windows version.
- How do I turn my voice into text on a Mac?
- Set the app up once — two permissions, a microphone check and a trial dictation — then press the right ⌘ key, speak and press it again (holding the key instead is an option in Settings): the words land where your cursor is. For a voice memo you already have as a file, bring it in as a dictation instead — turning an audio file into text.
- Does voice to text work in any app with a text field?
- In any app with a text field. Dalm puts the text into the field through Accessibility, so it needs that permission; if a field does not accept it, the text goes in through the clipboard and ⌘V, and the clipboard is restored after 300 ms. It has been checked in Telegram, Notion, TextEdit and Safari, including with a non-English keyboard layout. It keeps no list of apps: the text goes to whichever app is in front, a terminal or a code editor included, though those have not been checked by hand. If there is nowhere to type, the dictation is saved and you can copy it from the widget.
- Can I dictate a prompt to an AI chat?
- Yes: a chat box is a text field like any other. For a long prompt, press the key once to start and again to stop. Dalm inserts the text and formats it for the app you're typing in. Dalm does not send your dictation to a language model; the chat receives the text you put in its field, and the dictation itself is synced to our servers like any other.
- Does voice to text send my voice to the cloud?
- With Dalm, by default no audio is sent anywhere: the speech is recognised on your Mac, and the audio is deleted once the text is ready; if processing fails, it stays on your Mac. The text is synced to our servers automatically, with no off switch, and nobody else can see your dictations. 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.
- Is there a free voice to text app for Mac?
- Dalm dictation is part of the Free plan, which is a plan rather than a trial. You need a Mac with Apple Silicon · macOS 15.0 or later and an account; what the paid plans add is on the pricing page.