- Tutorial
- Voice to text
- Dictation
Voice to text in any Mac app, not just the chat box
How to turn speech into text wherever your cursor already is: built‑in macOS dictation, chat assistants, and a global hotkey that types into any app.
Most voice‑to‑text demos have a hidden catch: the words land in the assistant’s own window. You talk to a chat app, it writes text into a bubble, and then you copy it, switch apps, paste it, and fix the formatting. If the text has to end up in a ticket, a Slack reply, or a commit message, the assistant added steps instead of removing them.
The goal is simpler: speak, and the text appears where your cursor already is. Here are the three ways to get that on a Mac, and where each one breaks down.
Option 1: macOS dictation (already installed)
macOS has dictation built in, and it types straight into the focused field:
- Open System Settings → Keyboard and turn on Dictation.
- Note the shortcut. By default it’s pressing Fn (Globe) twice, or the Microphone key if your keyboard has one.
- Put the cursor in any text field, press the shortcut, and talk.
It’s free, it works in almost every app, and on Apple Silicon much of it runs on‑device. Where it falls short: no cleanup of filler words, no custom dictionary for the names and jargon you use every day, and punctuation that needs a second pass. I compared it with the paid options in more detail in my post on offline dictation for Mac.
Option 2: chat assistants (the copy‑paste dip)
Talking to a voice assistant is great for thinking out loud. It’s the wrong shape for filling in a form. The workflow looks like this: dictate into the assistant, copy the result, switch to the real app, paste, clean up, and delete the assistant window. That’s a pile of steps that didn’t exist when you typed by hand.
Chat assistants also can’t see the field you’re typing into, so they don’t know whether you’re writing a commit message, a bug description, or a chat reply — and each of those wants a different tone and length.
Option 3: a global hotkey that types anywhere
This is the shape I built into Snapbak. Hold a hotkey, speak, release, and the text is inserted into whatever app has focus:
- Slack replies, so a long answer takes seconds of talking instead of minutes of typing.
- Commit messages and code comments, straight into the editor.
- Ticket descriptions, written while the bug is still on screen.
- Long prompts for AI coding tools, where typing speed is the bottleneck.
A few things I cared about while building it:
- It runs on your Mac. The speech model (Parakeet) is on‑device, and dictation is English‑only. Nothing you say is sent to a server.
- A custom dictionary. Add product names and teammates’ names once and they come out right every time.
- Spoken commands. Saying “new line,” “comma,” and “period” out loud beats editing punctuation later.
- History. The last things you dictated are one click away when a paste goes wrong.
And because dictation is bundled with the capture app, it costs no extra menu bar icon and no extra subscription. The whole installer is targeted at ~15–20 MB.
Small habits that make dictation work
- Speak in sentences, not fragments. Cleanup is better on full sentences.
- Put names and code identifiers in the dictionary the first time one comes out wrong.
- For short utterances, hold the hotkey, talk, release; for long drafts, dictate a paragraph, fix it, then continue.
The hard places: terminals, numbers, and dates
Two dictation frontiers trip people up. The terminal works like any other app — the text lands where the cursor is — but shells care about characters prose doesn’t: say “quote dash dash flag quote” and you’ll type more than you save. The working pattern is to dictate the prose (commit messages, PR descriptions, documentation) and leave flag‑and‑punctuation‑heavy commands to the keyboard. Numbers and dates are the other frontier: speak them the way you want them written — “May twenty‑first” or “May 21” — because most engines write what they hear. The custom dictionary can’t fix a format — it teaches names and jargon — so the way you say it is the whole fix.
Which one should you use?
- Thinking out loud with an AI: the chat assistant, on purpose.
- Occasional short dictation, no new apps: macOS dictation.
- All‑day writing in real apps: a global‑hotkey tool. If it also records your screen and takes screenshots, that’s a bonus.
Voice‑to‑text stops feeling like a party trick the day it types into the app you were already using.