Dictation records while you hold its hotkey, turns your speech into text, applies your selected polish and writing style, and inserts the result at the current cursor.
Dictate into an app
- Click the text field where the result should appear.
- Hold your Dictation hotkey. The default is Fn.
- Speak naturally. The overlay shows that mrmr is listening and, when enabled, previews words as they arrive.
- Release the hotkey to finish.
Very short accidental taps are discarded. A recording warns at two minutes and stops automatically at five minutes.
You don’t have to wait for one dictation to clear before starting the next: hold the hotkey again while the result is still showing in the overlay.
The overlay
Nothing is on screen until you start. While you dictate, mrmr shows a small overlay: a glass capsule at the bottom of the screen, or a black drop from the notch if you chose that under General → Presence. Its orb is blue while it listens and turns lavender while your text is polished, then the overlay leaves as the text is inserted.
Double-click the overlay to open mrmr’s main window.
Speak naturally
You don’t need to dictate formatting the way you would type it. mrmr cleans up the transcript before inserting it:
- Corrections — change your mind mid-sentence and only the final version is kept: “Let’s meet Tuesday, no wait, Wednesday” becomes “Let’s meet Wednesday.” “Scratch that” and “I mean” work the same way. An emphatic “actually” (“I actually think…”) is kept.
- Punctuation — say “comma”, “period”, “question mark”, “new line”, or “new paragraph” where you want them.
- Lists — say “bullet points” or “numbered list” to get one. Saying “first… second… third…” stays as a sentence, except in notes apps, where a spoken list is formatted as one.
- Numbers — money, percentages, dates, times, versions, and quantities above ten are written as digits (“$20”, “15%”, “macOS 26”); small counts in a sentence stay as words.
- Questions and requests — a dictated question or instruction, even one addressed to an AI, is inserted as text. mrmr never answers it.
Dictating into the middle of text
When the app reports where your cursor is, mrmr fits the result to the text around it: it adds the space that separates it from the word before, continues a sentence you are in the middle of without a capital letter, and leaves off the period when the sentence carries on after the cursor. Dictating twice in a row without moving the cursor gets the space between the two results even in apps that don’t report the cursor.
In Slack, Notion, Discord, and similar apps built on Electron, mrmr can read the text around your cursor from your second dictation in that app onward. Code editors based on VS Code (VS Code, Cursor, Windsurf, Antigravity) are left alone, because they would switch into screen-reader mode.
Use hands-free mode
Use either method:
- Double-tap your Dictation hotkey.
- While holding the hotkey and dictating, tap Space.
The recording stays active after you release the hotkey. The overlay shows the icon of the app your text will go to, your words with a timer, and three buttons:
| Button | Shortcut | What it does |
|---|---|---|
| Cancel | Esc | Discards the recording |
| Start over | Shift + Esc | Clears what you have said and keeps listening |
| Submit | Your Dictation hotkey | Finishes and inserts the text |
If you move to another app while dictating hands-free, the text goes to the app in front when you finish.
If mrmr hears nothing for 10 seconds, hands-free ends with Didn’t hear anything and a Troubleshoot link to your microphone settings.
Cancel or start over
Press Esc while recording or processing. Cancelling prevents the current result from being inserted.
To clear what you have said without ending the recording, start over: press Fn + Delete while holding Fn, or Shift + Esc in hands-free mode. The microphone stays open and the timer restarts.
If you switch apps before the text is ready
When you hold the hotkey to dictate, mrmr pastes into the app you started in. If a different app is in front by the time the text is ready, mrmr follows Voice → Dictation mode → If you switch apps before the text is ready:
- Keep it on the clipboard (default) — nothing is pasted. The overlay shows You switched apps. Press ⌘V to paste. Hover over the notice to keep it on screen. Select Always paste on the notice to switch to the other behavior from then on.
- Paste into the new app — the text is pasted wherever you are now.
Either way, Auto-submit never presses Return in an app you switched to.
If text cannot be inserted
mrmr uses macOS Accessibility access to paste at your cursor. If insertion fails, the result is still saved: paste it with the Paste last transcript shortcut below.
Check General → Permissions → Accessibility if insertion repeatedly fails.
Paste the latest transcript again
Use the Paste last transcript shortcut shown under Voice → Dictation mode. The default is ⌘ + Control + V, and it can be changed.
Recording controls
Open Voice → Dictation mode to configure:
- Hotkey — the keys held to dictate.
- Live preview — show recognized words while speaking.
- Voice processing — noise suppression, echo cancellation, and automatic gain on your microphone. On by default; some microphones transcribe more accurately with it off.
- Mute while recording — silence other apps during dictation and restore them afterward.
- Auto-submit after dictation — press Return immediately after inserting text.
- If you switch apps before the text is ready — keep the text on the clipboard or paste it into the new app. See above.
- Paste last transcript — change the recovery shortcut.
The command palette (⌘ + K) can toggle Voice processing and switch the app-switch behavior too.
Auto-submit should be used carefully. It submits the polished result before you can review it, and Return inserts a newline rather than submitting in many editors.
Languages
Under Voice → Languages, select the languages you dictate in. Selecting one language locks recognition to it; selecting multiple keeps those languages available. mrmr preserves the language and script of your speech when polishing.
Audio feedback
Under Voice → Feedback, you can turn sound effects on or off, choose the Glass or Pulse sound style, set the sound volume, and enable or disable trackpad haptics. These settings do not change transcription behavior.
