For attorneys on Mac
Dictation that keeps client confidences on your Mac.
Hold a key, speak, release. Your words are typed into the document in front of you, and neither the audio nor the text ever leaves your computer.
No account · no cloud
Mac today · Windows planned
Thank you for meeting with me on Tuesday. This letter confirms the scope of my representation and the basis on which fees will be charged.
network connections: 0audio saved: none
Zero bytes uploadedRecognition runs on your computer.
Zero audio files savedSound is processed and gone.
Zero accounts to createNo login. No profile of you.
Why this exists
Dragon for Mac was discontinued in 2018. Most of what replaced it transcribes in someone else’s cloud.
For client matters that means a third party handles privileged material every time you speak. ABA Formal Opinion 512 (2024) is a reminder that the duty of confidentiality follows client information into AI tools.
Dictaquill takes the other road. There is no upload path, so there is nothing to disclose, nothing to get consent for and nothing to breach.
What it does today
Already working. In daily use.
- Hold, speak, release
- One key. Your words are typed where your cursor already is. No window to switch to, nothing to paste.
- Recognised on your Mac
- Speech recognition runs on-device with Apple's speech model. We measured the app and the recognition process while dictating: zero network connections.
- Audio is never saved
- Sound goes from the microphone to the recogniser and is gone. No recording is ever written to disk.
- Nothing is lost
- Dictaquill checks that your text actually arrived. If an app refuses it, the text waits on your clipboard and in your local history.
- Stops at password fields
- When a password field is active, dictation pauses. Nothing is typed and nothing is saved.
- History you control
- Kept on your Mac for 90 days, or not at all. Delete one entry or everything, whenever you like.
- Your names, spelled your way
- A personal correction list fixes the names and terms a general recogniser gets wrong.
- Stays off your other devices
- Text placed on the clipboard is marked local-only, so it does not sync to your iPhone or iPad.
NowMac. macOS 26 or later, built and tested on Apple Silicon.
PlannedWindows. Same approach: on your computer, nothing uploaded. Tell us in the form.
In development for attorneys
Not built yet. You decide the order.
Everything below is planned, none of it is finished, and we would rather say so. Tick what matters in the form and that is what gets built first.
- 01Legal vocabulary it actually knowsLatin terms, procedural language and reporter names recognised correctly, not corrected afterwards.
- 02Citation and section formattingSpoken citations written the way you would type them, by fixed rules rather than AI guesswork.
- 03Word lists per matterParties, judges and opposing counsel. Switch matter, and the right names are recognised.
- 04Learns from your correctionsFix a word once in your document and Dictaquill remembers, on your Mac only.
- 05Verified in Word and Outlook for MacTested typing into the two apps most firms live in.
- 06A confidentiality statement for your fileOne page stating exactly what the app does and does not transmit, and how to verify it.
You say
“pursuant to section nineteen eighty three plaintiff moves in limine to exclude the report”
It writes
Pursuant to § 1983, Plaintiff moves in limine to exclude the report.
Plain answers
Questions worth asking.
Does it work in Word and Outlook?
Dictaquill types into whichever app has your cursor. It is in daily use in browsers, text editors and terminals. Word and Outlook for Mac are being verified now. Tell us in the form if they are what you use.
Is anything sent to a server?
No. Recognition runs on your Mac and Dictaquill has no server that receives audio or text. When the paid version ships it will make two kinds of connection: a licence check and an update check. Neither carries anything you said, and both will be listed plainly in the app and the privacy policy.
Is my audio sent to Apple?
Dictaquill uses Apple's on-device speech model. On macOS 26.6 we inspected both the app and Apple's recognition process during dictation and found zero open network connections.
What do I need to run it?
Today, a Mac running macOS 26 or later; it is built and tested on Apple Silicon. A Windows version is planned, built the same way: recognition on your own computer, nothing uploaded. The form asks which you need, so we know what to build next.
What will it cost?
Not decided. The form asks what you would expect to pay, including “I wouldn’t pay for this”. An honest answer is the most useful thing you can give us.
Who makes it?
An independent developer. Joining early access puts you in direct contact with the person writing the code.