Never steals focus.
The pill floats above the Dock, on every Space, even over full-screen apps.
On-device dictation for Mac
Your fastest
keyboard is
your voice.
Speak, and OpenWhisper types clean, punctuated text at your cursor in any Mac app, about three times faster than your fingers. It all happens on your Mac.
Founding price. Rising to $249 soon. One payment, for Macs with Apple silicon.
Hold ⌥ Space to try it
⌥ Space
A simulation. The app runs on your Mac.
Types into
SlackMailXcodeTerminalNotesMessagesLinearNotionFigmaVS CodeCursorChatGPTObsidianGoogle DocsPages
01Faster than typing
You think at 150 words a minute. Your fingers manage 40.
Speaking is three to four times faster than typing. Hold a key, say it, let go, and it’s written, punctuated, in the app you’re already in.
Typing
~40 wpm
0:00
Speaking, with OpenWhisper
~150 wpm
0:00
Done. Punctuated.
2,000words
Typing, at 40 words a minute50 min
Speaking, at 150 words a minute13 min
37 minback, every day. About 142 hours a year.
Typical adult rates for conversational English, over 230 working days. The race runs at 5× speed; its clocks show real time.
How it works
Three moves. About a second.
- 01
Hold ⌥ Space
Or Right Option. A small pill floats above the Dock and never steals focus.
- 02
Speak
Talk normally. Say “comma” or “new paragraph” when you want them typed.
- 03
Release
Clean text is typed where your cursor is, in whatever app is in front. Your clipboard is left exactly as it was.
Cleanup
Say it the way you’d say it.
Spoken commands become real punctuation. Fillers vanish when you want them gone. Your names get spelled your way. All instantly, on your Mac. Try it: edit the left side.
What you said
What lands
Ready
Your words, spelled your way.
A custom dictionary fixes the names and jargon it mishears, every time. Add one and watch the right side.
Ums, gone.
Clean, Message and Code drop the fillers. Voice keeps every word you said.
umuhermhmmuhhumm
This demo runs the same rules as the app’s formatter, in your browser.
02Short or long
From a three-word reminder to a twenty-minute brain dump.
Hold the key for a quick reply. Double-tap it to go hands-free and talk for as long as the idea runs. Long takes show their progress, and every one is saved as text you can search, so nothing gets lost between thinking it and writing it down.
- Hold ⌥ Space for notes and replies that take seconds
- Double-tap for hands-free takes that run for minutes
- Search everything you’ve said, later, as text
0:00Reminder · Notes
Call Dana about Thursday.
4 words
0:00Reply · Slack
Love it. Ship the beta tonight and I’ll write the release notes first thing tomorrow.
16 words
0:00Email · Mail
Hi Sam, quick recap of today. We agreed to move the launch to Thursday, Priya owns onboarding, and I’ll have the pricing page ready by Wednesday. Two open questions before then: do we lead with privacy or speed, and who signs off on the founding price? I’d like answers by Monday so the team isn’t waiting on us.
325 words
0:00Brain dump · hands-free
Chapter three. The first time I tried to explain the product to my dad, I realised the problem was never the feature list, it was the order. People don’t care how it works until they believe it’s faster, and they don’t believe it’s faster until they hear themselves using it. So the chapter opens with the sound of someone talking, not with a diagram…
2,800 words
03Better answers from AI
Say more to AI. Get sharper ideas back.
Nobody types a hundred-word prompt. Everybody says one. Brief ChatGPT, Claude or Cursor the way you’d brief a colleague, with the background, the constraints and the why, and the answers come back focused on your actual problem.
Typed0 words of context
Spoken, with OpenWhisper0 words of context
An illustrative exchange. OpenWhisper types into any app with a text field, including the AI tools you already use.
Every app
One voice. Every app.
OpenWhisper types real keystrokes wherever your cursor is. It works in apps that block paste, and your clipboard is left exactly as it was.
Reply before the thought goes cold.
Mode Message
04Stays on your Mac
Your ideas stay on your Mac. Not on someone’s server.
Transcription runs on your Mac’s Neural Engine. The only network request OpenWhisper ever makes is the one-time speech model download. Switch on Airplane mode below and keep talking.
Network · OpenWhisper
Your microphone, on-deviceNetwork traffic
Sent while you dictate0 bytes
Model downloadOnce, at setup
StatusTranscribing on-device
Nothing to upload to.
Your audio, your text and the apps you use never leave your Mac. There is no server on the other end.
Works offline.
After the one-time model download it needs no connection at all. Planes, trains, basements.
No account. No telemetry.
It never signs in and never phones home. Your settings stay on your Mac.
05Built for Apple silicon
Transcribed by the chip already in your Mac.
OpenWhisper runs Whisper on Core ML and the Neural Engine inside Apple’s M-series chips. That’s why it’s fast with no internet, and why it needs an Apple silicon Mac.
Requires a Mac with Apple silicon
M1, M2, M3, M4, M5 or M6, running macOS 14 or later. It won’t run on Intel Macs, Windows or Linux.
Not sure? Apple menu → About This Mac → Chip: Apple M…
Whisper · Core ML · Apple Neural Engine · no network
Sized to your Mac
One download. Then it’s yours.
Six Whisper models run on the Neural Engine. On first run OpenWhisper recommends one sized to your Mac’s memory. Switch whenever you like.
Recommended for 16 GB: Small
Your Mac’s memory
06Your voice is yours
We don’t store your voice. We don’t train on it.
Your audio lives in memory for the second it takes to transcribe, then it’s gone. No recordings on disk, nothing uploaded, nothing used to train a model. What stays is your text, wherever you put it.
Move the launch to Thursday, and send Priya the deck.
- Recordings saved to diskNone
- Audio sent anywhere0 bytes
- Voice used for trainingNever
- Account to createNone
- What staysYour text, where you put it
Every good dictation app I tried wanted two things I was not willing to give: a subscription, and a copy of my voice. So this one does neither.
And the details
Built to disappear into your day.
A menu-bar app with no Dock icon to babysit. Six small things it does quietly, out of your way.
Follows your microphone.
Pick a mic or follow the system default. Plug in headphones mid-sentence and capture moves with you.
Type or paste, per app.
Real keystrokes by default, paste for the apps that block them. Your clipboard comes back either way.
SolidSpectrum
Make it yours.
26 accent colors and a Solid → Spectrum dial for the waveform. Go on, try it.
Speaks your language.
Set a language per mode, or let it auto-detect.
Silence, trimmed.
Quiet at the start and end of each take is cut first, for faster, cleaner results.
Made for
People who think out loud.
Same app, five very different days. Pick yours.
Engineers
Your hands are the slowest part of your stack.
Photographers
You shoot all day. Your inbox is cleared by dinner.
Coaches
Your programs should sound like you, not a keyboard.
Podcasters
You talk for a living. Stop typing the rest of it.
Project managers
Walk out of your last meeting with the notes already written.
07Pay once
One payment. No subscription. Ever.
Cloud dictation bills you every month, and your voice rides through someone else’s servers the whole time. OpenWhisper is one payment, then it’s yours.
OpenWhisper, onceCloud dictation at $15/mo
$0A monthly subscription
$0Cloud transcription minutes
noneAn account to manage
neverSomeone else holding your words
08Founding price
Get in at the founding price.
Early buyers pay less, once, and keep it for good. The price rises to $249 soon, and founding customers get every 0.x and 1.x update.
Founding price · rising soon
Lifetime license
For anyone who talks faster than they type.
$196.97once, not monthlyRises to $249 soon
- Lifetime license for up to 3 Macs you own
- Free updates through every 0.x and 1.x release
- 14-day free trial in the app. No account, no card
- Every feature, fully offline after one model download
Requires an Apple silicon Mac (M1–M6) running macOS 14 or later. Not for Intel Macs or Windows PCs.
FAQ
Questions people ask before buying.
Something else? Email the developer. One person reads that inbox.
Will it run on my Mac?
If your Mac has Apple silicon (M1, M2, M3, M4, M5 or M6) and runs macOS 14 or later, yes. It won’t run on Intel Macs, Windows or Linux. To check, open the Apple menu, choose About This Mac, and look for “Chip: Apple M…”.
Do you store or train on my voice?
No. Audio is processed in memory on your Mac and discarded the moment transcription finishes. Nothing is recorded to disk, uploaded, or used to train any model.
Is anything sent to the cloud?
No. Transcription runs on-device with WhisperKit on Core ML. The only network use is downloading a speech model the first time.
Is this a subscription?
No. One payment, yours outright, with free updates through every 0.x and 1.x release.
How long does the founding price last?
It’s the price for early buyers, and it rises to $249 soon. Buy at the founding price and that’s what you pay, once.
Can I dictate something long, like a whole chapter?
Yes. Double-tap ⌥ Space to go hands-free and talk for as long as you like. Long takes show their progress while they transcribe, and every dictation is saved as searchable text.
Does it work with ChatGPT, Claude and Cursor?
Yes. OpenWhisper types into any app with a text field, so you can speak a detailed prompt anywhere you’d type one.
Is there a free trial?
Yes. Every feature for 14 days, right in the app. There’s no account to create and no card to enter.
Why does it ask for Accessibility permission?
For the global shortcut and to type into other apps. If you decline, it still works and places the text on your clipboard instead.
How many Macs can I use it on?
Up to three that you personally use, on one license.
Your Mac is ready
Start talking to your Mac.
Free for 14 days. Then $196.97 once, at the founding price, before it rises to $249.