macOS 14+ · Apple Silicon · On-device by default · No account · No transcript history
Justfinishedthedraft.Icansenditoverafterlunch.
listens · finalizes · types
You sayWhisprTyper typesJust finished the draft. I can send it over after lunch.
WhisprTyper finalizes your words, then returns them to the supported editable field that was active when you started. If that field is unavailable, the transcript stays available in memory.
Works across your Mac
No plugins or per-app setup. WhisprTyper works with supported editable fields and intentionally refuses secure password fields.
Email
Chat & messages
Notes
Documents
Code editors
Browser text boxes
Supported editable fields
How it works
Hold. Talk. Let go.
⌃ control
Hold Control
Quick push-to-talk on the bare Control key is enabled by default. Hold it down from any app.
Talk
Say what you want to write, in your own words. The default engine transcribes on your Mac.
release
Let go
The finished words return to the supported editable field that was active when you started speaking.
Prefer a different trigger? The global shortcut defaults to ⌃⌥Space and is configurable, with push-to-talk, toggle, and automatic speech trigger modes.
Built to be trusted with your words
Transcription happens on your Mac
The default WhisperKit Multilingual engine runs locally after a one-time model download of about 626 MB. Once cached, it can work offline. No accounts, no ads, no analytics, no developer-operated transcription cloud.
✓WhisperKit Multilingualon this Mac · default
Apple System Speechoptional · may send audio to Apple
Pick how capture starts
Hold to talk, toggle it on, or let it start when you begin speaking.
Push-to-talk
Toggle
Speech trigger
Words return to the original field
WhisprTyper captures the supported editable field that is active when recording starts and attempts to insert only there. If the field cannot be safely recovered, it keeps the transcript in memory instead.
See you at sixinserted here
Password fields are refused
WhisprTyper will not type into password fields. Not by accident, not on request.
••••••••refused
One transcript, in memory only
If insertion fails, your latest transcript is held in RAM so you can copy it or retry the original destination. Retry is unavailable after successful insertion; memory clears when the app quits.
Copy Last Transcript
Retry Last Insertion
Quit clears memory
Lives in the menu bar
A native macOS 14+ app that waits quietly next to your clock. No Dock icon, no window in your way.
Say the whole reply instead of thumbing out a short one. It lands in the message box, ready to review and send.
“Sounds good, I can pick it up after lunch. Want me to bring the printouts?”
Documents
Dictate the rough first draft while the thinking is fresh, then shape it with your keyboard.
“The goal this quarter is to cut onboarding time in half. Here’s the plan.”
Support replies
Talking makes it cheap to be thorough. Walk through the fix out loud and send the complete answer.
“Thanks for flagging this. Here’s what’s happening and two ways to fix it.”
Code prose
Commit messages, review comments, TODOs. Dictate the words around your code and keep your hands on the problem.
“Fix race in session refresh: take the lock before reading the expiry timestamp.”
Privacy
Your words stay off our servers
WhisprTyper has no account system, no ads, no analytics, no tracking, and no developer-operated transcription backend. Audio is used for the current transcription; the only thing kept is one transcript in memory, for recovery, until you quit.
If you explicitly switch to the optional Apple System Speech engine, macOS may send audio to Apple under Apple’s privacy policy. The default engine stays on your Mac.
The current direct download requires an Apple Silicon Mac running macOS 14 (Sonoma) or later. It’s a Mac-only menu bar app; there is no Intel, Windows, iPhone, or iPad version.
Why is there a 626 MB download on first use?
The default WhisperKit Multilingual engine runs entirely on your Mac, so the speech model has to live there too. It’s a one-time download of about 626 MB, cached locally and reused.
Does it work offline?
Yes. Once the model is cached, the default engine transcribes on-device and can work without a network connection. Only the initial model download needs the internet.
What permissions does it need?
Microphone, to hear you, and Accessibility, to insert the finished text into the field you were using. Speech Recognition permission is requested only if you switch to the optional Apple System Speech engine.
What is the optional Apple System Speech engine?
You can choose macOS’s built-in speech recognition instead of WhisperKit. Depending on your macOS version and language, Apple’s engine may send audio to Apple’s servers, governed by Apple’s privacy policy. The default engine stays local.
What data does WhisprTyper collect about me?
None through the app: no account, no analytics, no tracking, no ads, and no transcript history. Your most recent transcript is held in memory only for recovery and clears when the app quits. Details are in the privacy policy.