Guides

Voice to text on a Mac, when it has to replace the keyboard

By Ryleigh Newman · Published September 14, 2026 · Updated October 6, 2026

Voice to text on a Mac starts with Apple Dictation, which is free and fine for a sentence at a time. For voice typing all day, YapToText types what you say into any app, shapes it for that app and runs entirely on the Mac, free, with no account.

My hands do not work well enough to type, so every email, message, document and line of code I write is spoken. This page is about what voice typing has to do to carry a whole day, and what YapToText does about each part. If you want a survey of every free option first, read my ranking of the best dictation software for Mac and come back.

Download for free. No account, nothing locked, macOS 14 or later on Apple silicon. Source on GitHub.

Steps

1. Install YapToText free from the Mac App Store on an Apple silicon Mac with macOS 14 or later.
2. Allow the microphone, then turn on the Accessibility switch in System Settings so the app can type into other apps.
3. On the Dictation page, in the Keys card, choose the dictation key and whether it taps to start and stop or holds to talk.
4. Put the cursor where the text belongs, tap the key, talk, and say question mark at the end of a question.
5. Tap the key again. The words are typed at the cursor, shaped for the app you are in.
6. In Settings, Advanced, Text output, use After a dictation lands to press Return for the apps where a message should go out on its own.

What a keyboard replacement has to do

A sentence here and there hides most problems. Dictating all day exposes them.

• Start and stop without holding anything. One key, tapped, or a switch that presses it.
• Keep up. A long dictation should not end in one long wait.
• Put the punctuation where you said it. With no stray mark left behind.
• Come out in the right shape. An email should read like an email and a chat message should stay casual, without you choosing.
• Land cleanly mid-sentence. With the right case and spacing.
• Send. A dictated message is not done until Return has been pressed.

Apple's built-in Dictation covers the start well. Its guide says it works anywhere you can type, takes text of any length without a timeout, and stops when it hears no speech for 30 seconds. Its command list covers punctuation, new lines, capitals, symbols and emoji, and nothing for selecting, deleting or replacing text already there. Try Apple Dictation first; everything below is about the gap after it.

Start and stop without holding a key

One tap of Right Command starts a dictation in any app, and one more tap ends it. Choose Hold to talk in the Dictation page's Keys card if you prefer a walkie-talkie: it records while the key is down and stops the moment you let go. Any single key can be the trigger, so anything that presses one key can run everything, though I have not tested every switch interface. If holding a key hurts, the carpal tunnel dictation page explains why a tap is the default.

While a dictation runs, Space pauses and resumes it and Esc cancels. Escape is only captured while you are dictating, and a canceled dictation is still transcribed and saved to History, marked as canceled, so a stray press cannot lose a paragraph. If your hands shake, turn on Require pressing Escape twice to cancel in the same card. A stray Esc stops the typing, not the saving.

Length is where a keyboard replacement usually falls over. Since 1.2, long recordings are cut at natural pauses instead of leaving you on a spinner. Now, once a dictation runs past 4 minutes, pieces of about 2 minutes are cut at quiet moments and transcribed while you are still talking, so the wait at the end covers only the last piece. If you speak quietly, Auto-amplify quiet speech measures your voice against the room instead of a fixed level. And when you would rather not touch a key at all, the menu bar panel's Start Dictation button does the same job as the key, which is how Voice Control can start a dictation by name. The keys and shortcuts page has every option.

Put the punctuation where you said it

Most sentences need nothing: Whisper punctuates ordinary speech on its own, and cleanup fixes the rest. Spoken names are for the marks it cannot guess, and since 1.4 they follow the standard dictation convention.

Say: End a question with the name of its mark.

is it working now, question mark

What happens: You get "is it working now?" with no stray mark left behind. Said in the middle of a sentence, a punctuation name stays a word.

Two-word names, such as question mark, exclamation point, new line and new paragraph, work on their own at the end of a clause. One-word names and emoji take the word "insert" first while Require the word insert before a command is on, which is the default, so an ordinary sentence never fires a command by accident. Symbols join the next word: "insert hashtag vibes" types #vibes. The full list is on the spoken punctuation page, and the Mac dictation commands page sets it beside Apple's.

For comparison, Apple's guide says Dictation adds commas, periods and question marks for you in supported languages, and that a spoken new line or new paragraph appears when you finish dictating. Google Docs voice typing adds punctuation through spoken commands, and it works only in Google Docs and in Slides speaker notes and captions, in Chrome, Edge or Safari.

Let the shape follow the app

The same words want a different shape in Mail, in Messages and in a code editor. Modes handle that: Raw Transcription types exactly what you said, Clean Up removes filler and false starts and fixes grammar, and Note, Email, Message and Code Comment each reshape your words for that kind of writing. Any app can have its own default mode, and the modes help page shows how to rewrite one in plain language.

I stopped picking modes when I built Auto mode, which is on by default. A letter with a greeting and a sign-off becomes an email, signed with your name if you gave the app one, casual chat stays exactly as spoken, and everything else gets a conservative pass that fixes what was misheard and rewrites nothing. The app counts too: Mail leans Email, Messages and Slack lean Message, and code editors lean Code Comment. Auto never turns speech into a list on its own. Select the message you are answering before you start, and your reply is written to fit it. Press a number key while you talk to pick a mode for that one dictation.

Say: End any dictation with this while Auto mode is on.

make that formal

What happens: The phrase is not typed. Auto mode rewrites the dictation in a formal tone. "As a bullet list" and "make it concise" work the same way, and the Auto mode page lists every phrase.

Other apps solve this their own way. Spokenly's documentation describes modes that turn on by themselves for a specific app or website, and superwhisper's Super Mode, part of its Pro plan, reads the active app, your selection and your clipboard.

App by app, with each app's own voice features side by side: Slack, Messages, Gmail, Notion, Obsidian, Discord and WhatsApp.

Dictate into the middle of a sentence

Text dropped into existing writing usually needs fixing afterwards: a capital that should be lowercase, a missing space, a period that closes a sentence you meant to continue. When your hands cannot do that fixing, every insertion is a second edit.

Intelligent Insert, on by default, reads the words around your cursor before the text lands. The first word is lowercased mid-sentence, with I and acronyms left alone, one space goes where words would collide, and a trailing period is dropped when the sentence carries on. Dictating over a selection replaces it, shaped like the selection. You dictate into the sentence you already have, and it fits.

The read uses a few invisible keystrokes, which is why some apps beep; the Intelligent Insert page has the cure, and you can turn the feature off for one app on the Dictation page. Terminals were the last thing to work well, and 1.5.1 fixed the part that hurt: in Terminal, iTerm2, Warp and the other terminals those keystrokes are shortcuts, so the read is skipped there and the words still land.

Send it without reaching for Return

A dictated message that sits in the box waiting for Return is not finished. In Settings, Advanced, Text output, After a dictation lands lets you choose, app by app, Return or Command Return, pressed the moment your words land, so a message or a prompt goes out on its own. The 1.5 release notes called it Send it for me. I keep it on for the apps where a message is always the whole job, and off where I want to read it first.

If you would rather see the text before anything lands, turn on Review before inserting in Settings, Advanced: finished text opens in a floating editor, Command Return inserts it and Esc discards it. Speaking is done when the message is sent, not when the words appear.

Fix words without going back to the keys

Right after text lands, say "scratch that" and it is taken back, or say "replace X with Y" and one word is fixed in place. No key is needed, only the exact phrase, said within 30 seconds. To teach a name the speech model keeps missing, select the right spelling, tap the Quick Edit key and say "add this to my dictionary"; dictionaries shape what the app hears, not just what it types, and the Dictionaries page has the rest.

For anything bigger, select the text in any app, tap the Quick Edit key, Right Option by default, say the change and tap again: make this shorter, fix the grammar, turn it into bullet points, translate it to Spanish. The result replaces your selection. VoiceInk also rewrites a selected passage when you say how it should change, and superwhisper's Super Mode handles formatting commands on selected text, per their own sites. What is different here is the price, which is nothing, and that all of it runs on your Mac.

How long the wait is

One number decides whether voice typing feels like a keyboard: the pause between stopping and seeing text. I am not going to promise you that number, because your Mac, your microphone and the length of what you said all change it, and a figure from my Mac would only be true on my Mac.

What I can give you is a way to measure your own. Every entry in History shows its stop-to-text time, split into transcription, AI cleanup and delivery, so you can see which step is slow, and whether a lighter model or a longer Keep models in memory setting on the Energy page helps. History shows where the time went on every dictation, so you can measure yours.

If voice typing stops keeping up

When a day of dictation starts to feel slow or wrong, climb this ladder, smallest step first:

1. Look at the wave. If it stays flat while you talk, the app is not hearing your microphone. Choose the right input in Settings, General, Microphone.
2. Wait out the first one. The first dictation after launch, or after the models unload, waits while the model loads. By default the models stay in memory for 2 minutes after use, and the Energy page can keep them longer.
3. Check where the words went. If they landed on the clipboard, Accessibility is off, or YapToText was the app in front.
4. Override the shape. If Auto picked the wrong format, press a number key while you talk to choose the mode for that dictation.
5. Read History. Every entry in History shows the raw transcript, the final text and the stop-to-text time, split into transcription, AI cleanup and delivery, so you can see which step was slow.

What it cannot do

• Clicking and moving around. YapToText writes; it does not click buttons or switch windows. macOS Voice Control does that part. If you run both, say "Command mode" so Voice Control answers only to commands and does not type your words as well. The guide to typing on a Mac without hands covers the split.
• Intel Macs. Nothing is claimed for them. It needs Apple silicon and macOS 14 or later.
• Speakers and timestamps. It types words. It does not label who said what or when.
• Languages beyond the list. The Language picker lists 29 languages plus Auto-detect, and it starts on your Mac's system language.

If you searched for talk to text rather than voice typing, the talk to text on a Mac page covers the same ground from the MacBook side.

The short version

• Short bursts. Apple Dictation is built in and enough.
• All day. YapToText taps to start, types into any app, shapes the text for that app and saves every dictation.
• Without the keyboard. Spoken punctuation, voice corrections, Quick Edit and a send key cover what used to need your hands.

Voice typing can carry a whole day when the app does the fixing your hands cannot. If something here does not match what you see, tell me on GitHub. Every issue gets read.

Download for free. No account, nothing locked, macOS 14 or later on Apple silicon. Source on GitHub.

About the author. Ryleigh Newman is a developer whose hands make typing difficult and who dictates every day with YapToText, the app this site describes. Facts about other apps come from their own sites, with the date they were read. More about the project.

Related

• Hands free typing on a Mac: the setup I use every day
• Auto mode: dictation that works out what it is
• Intelligent Insert: dictate into the middle of a sentence and it fits
• Speech to text for Mac that runs on the Mac
• Google dictation app for Mac: what Google offers, and the on-device alternatives
• Writing a book by voice on a Mac: dictation for writers

Sources

• https://support.apple.com/guide/mac-help/use-dictation-mh40584/mac (checked 2026-10-03)
• https://support.apple.com/guide/mac-help/commands-for-dictating-text-on-mac-mh40695/mac (checked 2026-10-03)
• https://support.apple.com/guide/mac-help/use-voice-control-commands-mh40719/mac (checked 2026-10-03)
• https://support.google.com/docs/answer/4492226 (checked 2026-10-03)
• https://spokenly.app/docs/modes (checked 2026-10-03)
• https://superwhisper.com/docs/modes/built-in (checked 2026-10-03)
• https://superwhisper.com/docs/billing/plans (checked 2026-10-03)
• https://tryvoiceink.com/ (checked 2026-10-03)
• https://github.com/ryleighnewman/YapToText (checked 2026-10-03)
• https://apps.apple.com/us/app/yaptotext/id6786382289 (checked 2026-10-03)

Questions