Private dictation for Windows & Mac

Dictation that never leaves your computer.

Talk into any app and get clean, punctuated text, transcribed on your own machine. Built for lawyers, clinicians and anyone under NDA.

  • $89 once, no subscription
  • 14-day free trial, no card
  • 14-day money-back promise

New message

To dana@example.com
Subject Re: Revised agreement

Hi Dana, thanks for sending the revised agreement. I've read it through and sections 2 and 4 look right to me. One change: can we push the delivery date to the 15th so there's time for the client's review?

  • Audio never uploaded
  • Works offline after setup
  • No account to create
  • One license, Windows and Mac
How it works

Hold a key. Talk. Let go.

  1. 1

    Hold the hotkey

    Put your cursor anywhere you can type. Hold Space on Windows, or fn on a Mac.

  2. 2

    Say what you mean

    Talk naturally. Pause, rephrase, say “scratch that”, keep going.

  3. 3

    Let go

    Clean, punctuated text lands where your cursor was. Your audio never left your computer.

Works in Outlook, Word, Gmail, Slack, Teams, Notion, Obsidian, VS Code and your browser. Anywhere you can type.

Features

Built to disappear into your day

No new editor to learn. ScrinkleScribe works inside the apps you already use.

Your speech stays on your machine

Transcription runs on a Whisper model on your own computer, in every configuration. Your audio is never sent anywhere.

One key, any app

Hold the hotkey and talk. Text lands wherever your cursor is: email, Slack, Word, your code editor, a browser form.

Clean text, not a raw transcript

A local AI pass adds punctuation and drops the ums, uhs and stutters. Say “new paragraph”, “bullet point” or “scratch that” and it does what you mean.

Fits the sentence you're in

Capitalization and spacing match the text around your cursor, and the writing style adapts to the app: email, chat or Markdown.

Edit by voice

Select some text, hold the command key and say how to change it. ScrinkleScribe rewrites the selection in place.

Your words, spelled right

Add names, product terms and jargon to your custom dictionary so they come out right the first time.

Privacy, exactly

Your audio is never sent anywhere

Speech is transcribed by a model on your own disk, on both platforms, in every configuration. No account, no analytics, no telemetry. Here is everything that does touch the internet.

Your voice goes from the microphone to the speech model and local cleanup on your computer, then into your app. The cloud is not involved.Your computerMicrophoneWhisper speech modelLocal AI cleanupYour appCloud: not involved

Check it yourself: after setup, switch off Wi-Fi and dictate. It keeps working.

  1. A license check

    When the app starts it checks your key with Polar, our payment provider: the key and an activation id (plus your computer's name when you first activate). Never audio or text. Offline, it keeps working for 14 days.

  2. One-time setup downloads

    Installing fetches the speech model and the components the app runs on. After that, dictation works with no internet connection.

  3. An optional online engine, off by default

    If you choose to use your own cloud AI key for text cleanup, your dictated text (and up to 60 characters before your cursor) goes to that provider. The app asks before the first one. Your audio never does.

Pricing

Pay once. Keep it.

One license, Windows and Mac, up to three computers.

ScrinkleScribe license

One person. Windows and Mac. Yours to keep.

$89one-time

14-day money-back promise. Key emailed right after checkout.

  • Transcribed on your computer; audio never uploaded
  • Works offline after setup
  • Windows and macOS
  • Up to 3 computers, swap any time
  • All 1.x updates included
  • No subscription, no account
  • Email support

A free 14-day trial opens at launch, no card required. Prices in USD; sales tax or VAT is calculated at checkout by Polar, our merchant of record.

Buying for a team? Email us and we'll sort out licenses for everyone.

Questions

Does it ever go online?

Only in the three ways the privacy policy lists. Setup downloads the speech model and the components the app runs on, once. When the app starts it checks your license with Polar, sending the key and an activation id (and your computer's name when you first activate a computer), never audio or text; offline, it keeps working for 14 days. And if you choose to switch on the optional online cleanup engine with your own API key, your dictated text and up to 60 characters before your cursor go to that provider. It is off by default.

How accurate is it?

It uses OpenAI's open Whisper models, running on your computer: large-v3-turbo on an NVIDIA graphics card or a Mac with 16 GB or more, and the smaller small.en model otherwise. Accuracy depends on your microphone, accent and vocabulary, so the free trial is the honest test. Add names and jargon to the custom dictionary to help it.

Is it a subscription?

No. You pay once and the license never expires. All 1.x updates are included. A future 2.0 would be a separate, optional upgrade.

How does the free trial work?

Get a free trial key, install the app and use the complete application for 14 days on up to three computers. No card required. When it ends, buy a license to keep going; anything you dictated stays yours.

What is your refund policy?

If it doesn't work for you, email us within 14 days of purchase for a full refund. No questions asked.

What hardware do I need?

On Windows: Windows 10 or 11, 64-bit, with 16 GB of RAM recommended. An NVIDIA graphics card lets it run the larger, more accurate speech model. On Mac: macOS 13 or later on Apple silicon (M1 or newer).

Do I need an account?

No. The app has no sign-up and no login. You get a license key by email and paste it in. If you ever need the key again, Polar's customer portal sends a one-time code to your purchase email.

See all 13 questions →

Give your keyboard a break

Try the full app free for 14 days, then pay once if it earns its place.

Already have a key? Download the app

Email me the linkLaunching Nov 17