How do I turn on dictation on my Mac?

Open System Settings → Keyboard → Dictation and turn it on. By default you start it by pressing the microphone key or a shortcut you choose, then speak in a text field where macOS Dictation is available.

Why does Mac dictation type my "um"s and skip punctuation?

Built-in dictation is a literal transcriber: it types what it hears. Automatic punctuation helps but is inconsistent, and filler words are kept. AI dictation apps rewrite speech into finished text instead.

Does dictation work in every Mac app?

Built-in dictation works where macOS accepts standard text input. Meosu tries safe direct insertion in the field you chose; if that is unavailable, it copies the result to the clipboard so you can paste it without dictating again. App behavior can vary.

How to dictate on your Mac: built-in vs. AI dictation

Your Mac has shipped with free dictation for years, and most people have never used it past the first accidental press of the microphone key. Here’s how to actually set it up, what it’s genuinely good at, where it falls over — and what the new generation of AI dictation apps does differently.

Setting up built-in macOS dictation

Open System Settings → Keyboard → Dictation and switch it on.
Pick your language and microphone.
Choose the shortcut — the default is pressing the 🎤 key (or Fn twice, depending on keyboard and settings).
Put your cursor in a text field that accepts built-in dictation, press the shortcut, and talk. Say punctuation out loud: “comma”, “period”, “new line”.

That’s it. It’s free, reasonably private (much of it runs on-device on modern Macs), and fine for short bursts.

Where built-in dictation is enough

Short messages: a one-line reply, a search query, a reminder.
Hands-busy moments: cooking, holding a baby, a wrist having a bad day.
Rough capture you intend to edit anyway.

If that’s your whole use case, stop here — you’re done, and it costs nothing.

Where it breaks down

Built-in dictation is a transcriber: its contract is to type what it heard. Real speech doesn’t survive that contract:

“so um can you send the… the deck by friday i wanna look at the numbers before the board call”

A transcriber types that sentence as spoken. Now you’re editing — moving the cursor, deleting fillers, adding punctuation, capitalizing — and the editing usually costs more than typing would have. This is the actual reason most people who try dictation abandon it within a week. It’s not the accuracy; it’s the cleanup.

What AI dictation apps do differently

The new category — Meosu and several worthy competitors — produces what you meant:

“Can you send the deck by Friday? I want to look at the numbers before the board call.”

Fillers dropped, restarts resolved, punctuation and casing handled, formatting adjusted for where the text is going. The good ones also handle technical content (“snake case max retries equals three” → max_retries = 3), learn your names and jargon through a dictionary, and offer an undo phrase (“scratch that”) so speaking freely is safe.

For app compatibility, the honest contract matters: Meosu tries safe direct insertion in the field you chose, and if that is unavailable it copies the result to the clipboard so you can paste it without dictating again.

The honest trade-off: most of these apps process audio on their servers, so you should read their privacy pages with the same attention you’d give any tool that hears you think. The questions that matter: is history opt-in, is audio retained, and can analytics carry your words. (Meosu’s answers: opt-in, discarded by default, structurally impossible.)

A fair way to decide

Spend one day on built-in dictation. If you never once wished the output were cleaner, keep it and pay nobody. If you spent the day editing transcripts, spend day two on an AI dictation app — Meosu is free on your Mac, with unlimited on-device dictation and no card — and compare how much of what you dictated shipped with zero edits. That number will make the decision for you.