How to Actually Dictate on a Mac (Not Just Turn It On)

We make a dictation app for Mac, so our support inbox is a decent proxy for what confuses people about voice-to-text on macOS. The single most common thread isn't a bug report about our software — it's someone who turned on the wrong built-in feature, or the right one but in the wrong place, and can't figure out why "dictation" isn't doing what they expected. That confusion is worth clearing up before anything else, because it's not really about EmberType. It's about Apple shipping two voice features that share a word.

The two kinds of "dictation" on your Mac

macOS actually has two separate voice features, and they live in two separate settings panes.

Dictation lives in Apple menu → System Settings → Keyboard → Dictation. It does one thing: it turns your speech into text inside whatever field is currently focused — a Pages document, a Slack message, a browser search box. Nothing more. It doesn't click buttons, doesn't open apps, doesn't move your cursor around the screen.

Voice Control lives in Apple menu → System Settings → Accessibility → Voice Control, and it's a different animal entirely. Per Apple's own accessibility guide, "with Voice Control, you can speak commands to navigate the desktop and apps, interact with what's on the screen, dictate and edit text, and more." It's whole-Mac control by voice, built for accessibility use, and it happens to include its own dictation and editing mode as one piece of a much bigger toolset.

Here's why that distinction matters in practice: if you turn on Voice Control expecting simple Dictation, your Mac starts listening for system commands constantly, which feels invasive and unpredictable if you just wanted to write an email hands-free. And if Dictation stops responding, the fix usually isn't in the Dictation pane at all — it's checking whether Voice Control got enabled somewhere and is competing for the microphone. We've walked through the specific failure modes in more depth in our troubleshooting guide for when Apple dictation stops working. For this piece, assume you want plain old speech-to-text, so everything below is about the Keyboard-settings Dictation, not Voice Control.

Setting up Dictation, step by step

The actual setup is short. Here's what Apple's own Mac User Guide lays out, with the parts people usually skim past called out:

1. Open System Settings → Keyboard → Dictation. Click Enable. macOS will ask whether you want to share your dictation audio with Apple to help improve Siri and Dictation — that's opt-in, and declining it doesn't turn the feature off.
2. Check the on-device toggle. The same pane lets you confirm whether "your voice inputs and transcripts for general text Dictation... are processed on your device and not sent to Siri servers." This is worth actually looking at rather than assuming — on Apple Silicon Macs it's typically on-device once the language model finishes downloading, but Apple frames it as something to verify per setup, not a blanket guarantee, which is why the checkbox exists at all.
3. Set how you'll trigger it. If your keyboard has a physical Microphone key in the function row, that starts Dictation directly. Most current MacBooks don't ship with one, so there's no fixed default shortcut — you assign your own under Dictation Shortcut (Apple's own example is Option-Z), or you can always reach it from the Edit → Start Dictation menu in any app.
4. Pick your language(s). You can add more than one and switch between them mid-session, though Apple notes "Dictation is not available in all languages or regions, and features may vary" — worth checking before you build a workflow around a specific language.

Once it's on, you talk and text appears wherever your cursor is. Press Escape, or click the Done button that appears near your text, to stop.

The punctuation and formatting commands nobody explains well

This is the part that makes people give up on built-in Dictation after one try, because it isn't obvious you have to say punctuation out loud. Apple's guidance is literally "say the name of the punctuation mark, such as 'exclamation mark.'" So to write "Wait, really?" you say "wait comma really question mark." A few that are easy to miss:

• Say "new line" for a single line break (one Return).
• Say "new paragraph" for a paragraph break (two Returns).
• In supported languages, macOS can auto-insert commas, periods, and question marks based on your speech pattern without you saying them — and you can turn that off in the same Dictation settings if it guesses wrong too often.

None of this is hidden exactly, but it's buried deep enough in Apple's docs that most people just never find it, dictate a run-on paragraph with no punctuation, and conclude dictation "doesn't really work." It works — you just have to talk like you're narrating a telegram.

Where built-in Dictation quietly stops

There's a real piece of history behind the "Mac dictation cuts you off" reputation, and it's worth separating fact from folklore. In OS X Mavericks and Yosemite, plain Dictation genuinely was capped at roughly 30 to 40 seconds per session unless you turned on Enhanced Dictation, a separate download that ran speech recognition locally instead of round-tripping to Apple's servers and removed the timeout. That's not a rumor — it's how the feature actually behaved for years, and it's why so many old forum threads and "fix" articles still talk about a hard 30-second wall.

On current macOS, that specific limit is gone. Apple's guide states plainly: "you can dictate text of any length without a timeout." The 30-second number that survives today is different — Dictation "stops automatically when no speech is detected for 30 seconds," which is an idle cutoff, not a length cap. Pause to think for half a minute mid-sentence and it'll close the session; talk continuously for five minutes and it won't stop you. That distinction matters and most articles about this topic still get it wrong by quoting the old Mavericks-era limit as if it still applies.

What's actually missing isn't length. It's the things around length. There's no custom dictionary — dictate a client's name, a technical term, or your own product name repeatedly, and Apple's Dictation will keep mishearing it the same way every time, with no way to teach it. There's no AI cleanup pass, so filler words, false starts, and "um, actually, let me rephrase that" all land in the text exactly as spoken. And because it's built around individual text fields, focus quirks are common — click away mid-sentence, or dictate into a field that briefly loses focus during an app transition, and you can lose the session.

When Apple's free tool is genuinely the right call

I want to be fair here, because it would be easy to spend this whole section talking our own book. If you're dictating a text message, a search query, a two-line email — built-in Dictation is fast, it's already on your Mac, it costs nothing, and for short bursts the "no custom dictionary" and "no AI cleanup" gaps basically never come up. There's no reason to install anything for that use case. Apple built a genuinely useful tool for quick, casual dictation, and most people's daily voice-to-text needs are exactly that: quick and casual.

Where we built EmberType to pick up the slack

The reason we started building EmberType wasn't that Apple's Dictation is bad — it's that it's scoped for short bursts, and a lot of real work isn't short. When we talked to people dictating long-form drafts, technical documentation, or client notes full of names and jargon, the same three gaps kept showing up: no way to teach the system your vocabulary, no formatting pass to turn a rambling voice memo into clean paragraphs, and no consistency across apps that don't play nicely with system-level dictation focus.

So EmberType runs entirely offline on Apple Silicon using Whisper, adds a real custom dictionary for names and terms you say often, applies optional AI formatting to clean up filler and structure your paragraphs, and works the same way in every app on your Mac rather than field-by-field. None of that makes Apple's Dictation obsolete for a quick note — it makes EmberType the tool for the dictation sessions that actually run long, which is where we've found the built-in tool's limits start to show.

FAQ

Is Dictation the same thing as Voice Control on Mac?

No, and this is the single most common source of "dictation isn't working" confusion. Dictation, in System Settings → Keyboard, only converts speech into text inside whatever field is focused. Voice Control, in System Settings → Accessibility, lets you operate the entire Mac by voice — opening apps, clicking buttons, navigating the screen — and it also has its own text dictation and editing mode. They're separate features in separate settings panes, and only Voice Control controls the whole system.

Does Mac dictation still cut off after 30 seconds?

Not the way it used to. In OS X Mavericks and Yosemite, basic Dictation really did cap you at roughly 30 to 40 seconds unless you enabled Enhanced Dictation, which downloaded an on-device model to remove the limit. On current macOS, Apple's own documentation states you can dictate text of any length without a timeout — the only 30-second behavior left is that Dictation stops listening after 30 seconds of silence, which is an idle cutoff, not a length cap.

What is the keyboard shortcut for Dictation on Mac?

It depends on your keyboard. If your Mac has a physical Microphone key in the function row, that starts Dictation directly. Most modern MacBooks don't, so Apple ships no fixed default — you set your own in System Settings → Keyboard → Dictation Shortcut, or trigger it any time from the Edit → Start Dictation menu.

Does macOS Dictation work offline?

Apple's Keyboard settings let you check whether your voice inputs and transcripts for general text Dictation are processed on your device rather than sent to Siri's servers, and on Apple Silicon Macs that's typically the default once the language model has downloaded. Apple doesn't frame it as an absolute guarantee for every language and setting, though — the settings pane exists specifically so you can verify it for your own configuration.