The short answer Open System Settings → Keyboard → Dictation, toggle it on, then press the dictation shortcut in any text field and talk — saying punctuation out loud ("comma", "new paragraph"). That's genuinely it. This guide covers the details Apple's one-liner skips, the failure modes everyone hits, and the point where a push-to-talk Whisper app becomes the better tool.

Your Mac has shipped with a perfectly serviceable dictation feature for years, and most people have never pressed the key. Fair enough — nobody told them, and the discoverability is terrible. Here's the two-minute setup, the commands worth memorizing, and an honest account of where built-in dictation runs out of road.

Step 1: Turn it on

  1. Open System Settings (the menu → System Settings).
  2. Click Keyboard in the sidebar and scroll to the Dictation section.
  3. Toggle Dictation on and confirm the prompt.
  4. While you're there, check three things: the language, the microphone source (your Mac's internal mic is fine; a headset mic is better), and the shortcut — the key you'll press to start dictating. On recent Macs it's often the microphone key or pressing the 🌐 Globe key; you can change it right in this pane.

That's the whole installation. No download, no account — it's already part of macOS, and on Apple Silicon Macs many languages are processed on-device.

Step 2: Actually dictate

Click into any text field — Notes is a good sandbox — press your shortcut, and a small microphone indicator appears. Talk in complete phrases rather than word… by… word; every speech engine, Apple's included, uses surrounding context to decide between "their" and "there". Press the shortcut again, or Escape, when you're done.

Punctuation is on you. The commands that matter:

Say thisGet this
"period" / "comma". ,
"question mark" / "exclamation point"? !
"new line"a line break
"new paragraph"a blank line and a fresh paragraph
"open quote" … "close quote""…"
"cap" before a wordCapitalizes That Word
"numeral five"5 instead of five

macOS can insert some punctuation automatically these days, but the results are inconsistent enough that practiced dictators still speak it. It feels absurd for about ten minutes ("Hi Sarah comma new paragraph") and then becomes muscle memory you stop hearing yourself say.

Step 3: Fix the usual problems

  • Nothing happens when you press the shortcut. Re-check the shortcut in the Dictation pane — another app may own that key. Assign a different one.
  • It types the wrong language. Dictation language is separate from your system language. Set it explicitly in the same settings pane.
  • Terrible accuracy. Usually the mic, not you. If you're across the room from an open laptop, the internal mic is hearing your ceiling. Sit normally or use a headset, and kill loud background noise where you can.
  • It stops listening mid-thought. Built-in dictation goes idle after a stretch of silence. There is no setting for this; it's the design. Which brings us to the next section.

Where built-in dictation runs out of road

None of these are bugs. They're the shape of the feature — fine for a text message, increasingly costly for real writing.

  • The silence timeout. Thinking is silent. Pause to compose your next sentence and dictation may have clocked out before you resume — the classic "talked for ten seconds into a dead mic" experience.
  • Your vocabulary doesn't exist. Colleagues' names, product names, technical jargon — the engine guesses, wrongly, and there's no proper custom dictionary to teach it. "Kubernetes" becomes an adventure.
  • Raw output only. Every "um", every false start, every run-on lands in your document verbatim. For email that means a second editing pass, which quietly cancels the speed you gained.
  • It's a mode, not a key. You toggle in, you toggle out, you glance to confirm the indicator caught it. Push-to-talk — mic live exactly while a key is held — removes that entire class of uncertainty.

The upgrade: push-to-talk with Whisper

This is the part where we mention we make one of these, so weigh accordingly: OpenVoiceFlow is a free, open-source push-to-talk dictation app for macOS. The differences map exactly onto the limits above:

  • Hold to talk. Hold fn (or any key you pick), speak as long as you like, pause as long as you like — the mic is live while the key is down, full stop. Release, and text pastes at your cursor.
  • Whisper accuracy, on-device. OpenAI's Whisper models run locally via WhisperKit, and punctuation comes free from sentence structure — no "comma" required. Audio never leaves your Mac. (Which model? We wrote it up.)
  • A real dictionary. Teach it "Anaïs", "kubectl", or your product's codename once; it stops getting them wrong.
  • Optional cleanup. Off by default. On, it fixes grammar and drops filler words using your own API key — or a fully local model via Ollama.

Setup is a five-minute affair: download the DMG, drag to Applications, grant three permissions, let it fetch a model once. And it costs nothing, which we consider a feature and a principle.

A fair decision rule

If you dictate a few short bursts a week, built-in dictation is honestly enough — enjoy it, you're done, this was a short article. If you're composing email, documents, or AI prompts by voice daily, the timeout, the vocabulary gap, and the raw output will each tax you a little every single day, and a purpose-built tool pays for itself immediately — especially when the purpose-built tool is free.

FAQ

How do I turn on dictation on my Mac?

System Settings → Keyboard → scroll to Dictation → toggle on. Press the shortcut shown there in any text field and speak.

Why does dictation stop after a few seconds?

Built-in dictation idles out after silence. Push-to-talk apps sidestep this: the mic is on exactly while you hold the key, however long you think between sentences.

Can my Mac dictate offline?

On Apple Silicon, built-in dictation handles many languages on-device. Whisper apps transcribe entirely on-device — here's the full offline picture, tested with the Wi-Fi actually off.

Does dictation work in every app?

Anywhere there's a text cursor: Mail, Slack, browsers, terminals, IDEs. Both built-in dictation and push-to-talk tools type into whatever field has focus.

Ready for the push-to-talk upgrade?

Free, open source, on-device. Hold a key, talk, release — that's the entire learning curve.

More from the blog