Practical guide · about 13 minutes
How to use voice to text on Mac — in any app.
macOS has Dictation built in: turn it on in Keyboard settings and press the Microphone key or your shortcut. For a review step, offline model choice and meeting transcripts, VoiceToText adds a shortcut that pastes into any app.
Built into macOS
Turn on Mac Dictation and learn the shortcut.
Every Mac has Dictation built in, so there is nothing to install: you can speak anywhere you can type. These steps follow Apple’s Mac User Guide.
Turn on Dictation
Choose Apple menu → System Settings, click Keyboard in the sidebar, go to Dictation and turn it on, then click Enable. If macOS asks whether to share audio recordings to improve Siri and Dictation, choose Share Audio Recordings or Not Now.
Click where the text goes, then start
Start Dictation in any of these ways:
- Microphone key🎤
- The Microphone key in the top row, if your keyboard has one
- Function or GlobefnDD
- Start or stop dictation, from Apple’s list of Mac keyboard shortcuts
- Your shortcut
- The one set in Keyboard → Dictation → Shortcut
- Edit menu
- Edit → Start Dictation
Speak when the cursor pulses or you hear the tone.
Say punctuation and line breaks
In supported languages, Dictation inserts commas, periods and question marks for you. You can also say a mark such as “exclamation mark”, an emoji name such as “heart emoji”, or “new line” and “new paragraph”. Apple’s list of dictation commands adds symbols and capitals. On a Mac with Apple silicon you can keep typing while you speak.
Stop
Press EscapeEsc, the Microphone key, or your shortcut again. There’s no length limit, but Dictation stops on its own after 30 seconds without speech. If a word comes out underlined in blue, click it to choose an alternative.
Change the shortcut
In System Settings → Keyboard → Dictation, open the Shortcut pop-up menu to see which shortcut your Mac uses, or pick another. On Macs with a Fn (🌐) key, one option is pressing that key twice. For a combination that isn’t listed, choose Customize and press it, for example Option-Z.
To control the whole Mac by voice, with commands like “Click Done”, turn on Voice Control in System Settings → Accessibility; while it’s on, you dictate through Voice Control instead. Coming from Dragon? Read Mac dictation vs Dragon.
When an app helps
Built-in Dictation covers quick text. An app adds three things.
Apple’s Dictation is free, starts with a key press, and takes punctuation by voice. VoiceToText is a free app for the jobs Dictation isn’t built for. It has no voice commands or spoken punctuation, so keep Dictation for those.
Review before it’s pasted
The whole transcript opens in a panel first, so you can fix names, numbers and negations before anything reaches the app, or add another take with ⌘R.
A choice of models
Six local models work with the network off after one download: Parakeet for 25 European languages and Whisper for 99. Nine optional cloud models use your own key.
Meetings and files
Record a call’s microphone and system audio with no bot, or transcribe an audio or video file, then search all of it in History.
For the full comparison, read Apple Dictation vs VoiceToText. To try the app, follow the five steps below.
VoiceToText setup
Set up VoiceToText in five steps.
VoiceToText works system-wide rather than inside one editor. Once the permissions are granted, the same shortcut works in nearly any app that accepts a paste.
Install VoiceToText
Download the DMG, open it, and drag VoiceToText into your Applications folder. The app is signed and notarized for macOS. On first launch it asks for microphone access and starts downloading the default model.
Grant two permissions
Microphone lets the app hear you. Accessibility lets it start recording from the global shortcut and paste the finished text where your cursor is; macOS asks for it the first time you dictate. You can revoke either permission in System Settings.
Put the cursor anywhere and press Option plus Space⌥Space
Open Mail, Notes, Slack, a browser, a terminal, or any other Mac app. Click where you want the text, press the shortcut, and speak naturally.
Watch the recording card
A small floating card appears with a live level meter and a timer, without taking focus from your app. Press the shortcut again or click Finish when you’re done, or press EscapeEsc to throw the take away. If the model is still downloading or loading, the card shows its progress first. By default it’s a pill with round ✕ and ✓ buttons for Cancel and Finish; Settings → General → Recording HUD makes it Medium or Large.
Review, then paste
The text opens in a review panel where you can edit it before anything is pasted:
- ReturnReturn
- Paste at the cursor
- Shift⇧ReturnReturn
- New line
- Option plus Space⌥Space
- Your shortcut again also pastes
- Command⌘RR
- Resume: record another take, inserted at the caret
- Command 1⌘1–Command 9⌘9
- Run an AI action, if you’ve turned any on
- Undo button
- Appears after an AI action; steps back one action at a time
- EscapeEsc
- Cancel and discard the take
VoiceToText pastes rather than types: it saves your clipboard, puts the text on it, presses ⌘V for you, and puts your previous clipboard back a moment later. Turn review off and the text pastes as soon as it’s ready.
Mac permissions
Grant only the permissions your shortcut needs.
The default setup needs Microphone and Accessibility. Input Monitoring is requested only when you choose Right Control as a standalone shortcut. Recording meetings adds Screen Recording, which is used for system audio only.
Microphone
Required to record your voice. Enable VoiceToText in System Settings → Privacy & Security → Microphone; you can revoke access at any time.
Accessibility
Required to start dictation, for Esc to cancel, and to paste the transcript into another app. VoiceToText does not use it to read your screen or log keystrokes.
Input Monitoring, optional
Only the standalone Right Control shortcut needs Input Monitoring. The default Option+Space shortcut does not require this third permission.
Models & languages
Choose an offline speech-to-text model for your language and Mac.
VoiceToText has six local models. Each downloads once, then transcribes on your Mac without sending audio anywhere, even with the network off. There is no language setting, so pick a model that covers the language you speak.
Parakeet TDT v3, the default
The recommended local model and the one downloaded first (~470 MB). It covers English and 24 other European languages and recognizes which one you’re speaking. There is no language to set.
Whisper Large v3 and Turbo
Local Whisper models cover 99 languages and recognize which one you’re speaking. Large v3 has the highest quality score of the local models (8.0) but takes longer; Turbo (7.5) is quicker.
Whisper Small, Base, and Tiny
Smaller downloads, from roughly 250 MB down to under 50 MB, for Macs short on disk space. They make noticeably more mistakes, and they cover the same 99 languages.
Cloud models, with your key
Nine optional OpenAI and ElevenLabs models detect the language on their side (90 to 99+ languages). Four are live models, and with three of them text appears while you speak in the Large recording HUD, word by word or phrase by phrase. Audio goes to the provider, billed to your key at $0.18/hr to $1.02/hr.
Deciding between the two local families? Read Whisper vs Parakeet on Mac. For a privacy walkthrough, see how offline speech to text works on Mac.
Make it yours
Choose the dictation flow that feels natural.
Start with the defaults, then change only what removes friction from your own workflow.
Toggle or hold to talk
“Press to toggle” is the default: press once to start, again to stop. In Settings → Shortcut, switch Recording mode to “Hold to record” to talk only while the key is down. “Esc cancels dictation” is on by default; turn it off there if Esc should reach the app you’re in.
Pick a shortcut that suits you
Record any key with at least one modifier (⌘ ⌥ ⌃ ⇧), a function key from F1 to F20 on its own, or Right Control on its own, which needs Input Monitoring. A second, optional Conversation shortcut starts and stops meeting recording.
Review or instant paste
“Review before pasting” is on by default, so you can edit, resume, or cancel before anything lands. Turn it off in Settings → General and the text pastes the moment transcription finishes. A take that lost audio when the microphone dropped out still opens review so you can check it.
AI Actions, when you want them
6 built-in actions (Clean transcript, To English, Improve prompt, Fix grammar, Summarize, Essentials only) plus your own, each a name and an instruction. All are off by default: turn them on in Settings → Actions and add an OpenAI key in Cloud. They run from the review panel, so keep Review before pasting on to use them. Running one sends the transcript text to OpenAI (gpt-5.5), billed to your key.
Run it from the menu bar
The menu bar item starts and stops dictation and meeting recording. In Settings → General, turn off Show in Dock to keep VoiceToText in the menu bar only, pick System, Light, or Dark appearance, and choose whether it opens at login (it’s on after the first launch).
Trigger it from other tools
Raycast, Shortcuts, Stream Deck, or a script can open voicetotext://toggle, start, stop, or cancel. In Terminal, open -g voicetotext://toggle starts dictation without bringing VoiceToText forward, so the text lands in the app you’re using. The URL scheme controls dictation only.
Keep, search, or skip history
Each dictation is saved on this Mac with its audio and transcript. In History you can search transcripts, star favorites, play audio back, and re-transcribe with another model. Turn off Save recordings to stop keeping dictations. History keeps your newest 200 dictations; favorites, meetings and recordings with a summary or action items are never removed automatically.
Compare models in the Models pane
The Models pane lists all 15 models, 6 on your Mac and 9 in the cloud, with a 1–10 quality score based on published third-party word error rates and the hourly price of each cloud model. Filter by On this Mac or Cloud and sort by quality.
Use it anywhere
Dictate into Mail, Notes, Slack, browsers, and coding tools.
The focused text field receives the transcript. That includes documents, chat apps, browser search boxes, terminals, ChatGPT, Claude Code, Cursor, and nearly any native or web app that accepts a paste.
- Notes
- Slack
- Notion
- Google Docs
- ChatGPT
- Cursor
- Terminal
Need longer-form capture? VoiceToText can also record and transcribe meetings on your Mac.
Troubleshooting
Fix common Mac dictation setup problems.
Check permissions and the selected model first; most shortcut, paste, language, and offline issues come from one of those settings.
The shortcut does nothing
Confirm VoiceToText is enabled under Accessibility; Settings → General → Permissions shows what’s missing. If another app owns the same combination, record a different shortcut in Settings → Shortcut.
Right Control does nothing
The standalone Right Control shortcut needs Input Monitoring in addition to Accessibility. Enable VoiceToText under System Settings → Privacy & Security → Input Monitoring.
Nothing was pasted
Click into a text field first and check that Accessibility is enabled. The text arrives as a paste, so fields that block pasting, such as some password fields, won’t receive it. Unless you turned off Save recordings, the transcript is still in History, where you can copy it.
The model is still downloading
The default model starts downloading when VoiceToText first opens; other local models download from the Models pane. If you dictate before one is ready, the recording card shows download and loading progress. The download needs an internet connection once; after that the model works offline. Press Esc to stop waiting.
The words come out in the wrong language
There is no language setting. Parakeet covers 25 European languages automatically. For other languages, choose a Whisper model (99 languages) or an OpenAI or ElevenLabs cloud model; each detects the language on its own. Whisper is most accurate in English and other widely spoken languages.
Saying “comma” or “new line” types the word
VoiceToText has no voice commands. The model adds punctuation from how you speak, so words like “comma”, “period”, or “new line” are typed as words, and filler words are kept. Press Shift+Return in the review panel for a line break, or turn on the Clean transcript or Fix grammar AI action in Settings → Actions (it needs your OpenAI key), which turns spoken cues like “comma” and “new line” into formatting.
Transcription is slow or inaccurate
Try Parakeet for local speed, or Whisper Large v3 for the most accurate transcription on your Mac. Speak close to the microphone and cut background noise. VoiceToText records from the macOS default input, which you choose in System Settings → Sound.
My AirPods switched mid-sentence
When the microphone changes in the middle of a dictation, capture restarts on its own and the take continues. If it can’t, what was already captured is still transcribed, and the review panel tells you if audio dropped out for a second or more.
Transcription failed
The failure card keeps your audio. Retry runs the same recording again. If a cloud model has no key, Add API Key opens Settings → Cloud; for a permission problem, Open Settings takes you to the right place.
“Recording too short” or “No speech detected”
Takes under half a second, or with no speech in them, are dropped so the model doesn’t invent text. Watch the level meter on the recording card: if it barely moves, check the input device in System Settings → Sound.
For model, privacy, and language details, read the complete VoiceToText FAQ.
FAQ
Voice to text on Mac: common questions.
How do I turn on voice to text on Mac?
What is the keyboard shortcut for voice to text on Mac?
Does voice to text on Mac work offline?
How do I add punctuation when dictating on Mac?
Can I use voice to text in any app on Mac?
Ready to speak
Try voice to text in the app you already use.
Free, local by default, and source available on GitHub. Download the DMG and dictate your first sentence in a few minutes.