Secretary42 app icon Secretary42

Speak instead of type —
right where you work.

Secretary42 is a local-first desktop dictation app. Press a global hotkey, speak, and your words are transcribed on your device and pasted straight into whatever you’re writing.

Now available macOS (Apple Silicon) & Windows x64 On-device transcription

Press a hotkey. Speak. Paste anywhere.

Stay in the document, email, or editor you’re already in. Secretary42 turns speech into text on your machine and drops it at your cursor.

How it works

From thought to text in four steps

No app switching, no copy-paste dance. Dictation happens where you’re working.

  1. 1

    Press your hotkey

    A small overlay starts listening. The default is the backtick key (`); pick your own anytime.

  2. 2

    Speak naturally

    Talk the way you’d talk. Pause, think, and keep going — no special phrasing required.

  3. 3

    Transcribed on your device

    The bundled whisper.cpp engine turns your audio into text locally — no cloud transcription.

  4. 4

    Text appears where you work

    Your words are pasted at the cursor, or sent to the clipboard or a file — your choice.

Pricing

Simple pricing for local desktop dictation

Choose monthly or yearly billing.

Monthly
$9.99 / month

Billed monthly. Manage billing through Paddle.

  • Local on-device transcription
  • Choice of Whisper models, from compact to high-quality
  • Local transcription history
  • Product updates while your subscription is active
  • macOS (Apple Silicon) & Windows x64
Buy monthly

Buy through Paddle. Your license is delivered by email; paste the license token into Secretary42 to activate the app.

Checkout and billing are handled by Paddle. License activation is separate from the local dictation path. By purchasing you agree to the Terms; see the Privacy Notice and Refund & Withdrawal Policy.

Privacy architecture

Your dictation stays on your device

Audio and transcript text are processed locally by the bundled whisper.cpp engine. Secretary42 does not use cloud transcription, and there is no cloud transcription fallback.

On device Stays on your machine

  • Microphone audio
  • Transcribed text
  • Your local transcription history
  • App settings and your license

Online May use online services

  • Paddle checkout & license delivery
  • Whisper model downloads
  • App update checks & downloads
  • Support requests
  • Diagnostics — opt-in, off by default

Diagnostics are opt-in and off by default. If you turn them on, only sanitized crash and performance metadata is shared — never your audio or transcript text. Paddle checkout, license delivery, model downloads, updates, and support are separate online flows, not part of the local dictation path.

Features

Built for fast, local dictation

Global hotkey

Start and stop dictation from any app with a customizable key.

On-device whisper.cpp

Speech-to-text runs locally with quantized Whisper models.

Bundled engine & FFmpeg

No Python and no separate FFmpeg install — it ships with the app.

Local model management

Download, select, and remove Whisper models from the control panel.

Local transcription history

Past transcriptions are stored locally in SQLite — view, copy, or delete.

Automatic paste

Text lands at your cursor; or route it to the clipboard or a file.

Choice of models

Trade speed for quality — from a compact model to large, high-quality ones.

macOS Globe key

Optional Fn / Globe-key trigger on macOS for hands-on-keys dictation.

Models

Pick your speed-and-quality trade-off

Models download on demand and are stored locally. Small (Balanced) is the recommended default — a good mix of speed and quality.

XSTiny — fastest, lower accuracy~31 MB
SBase — fast with better accuracy~57 MB
MSmall Recommended — best balance of speed and quality~182 MB
LMedium — higher accuracy, slower~515 MB
XLLarge v3 — maximum accuracy, slowest~1.5 GB

Sizes are approximate download sizes for the bundled quantized models. Larger models need more disk space and processing time.

Use cases

Dictate the writing you do every day

Email replies

Clear answers and follow-ups without typing them out.

Notes & docs

Capture thoughts and meeting notes while they’re fresh.

Documentation

Draft internal docs and READMEs at the speed of speech.

Drafting

Get a messy first draft down, then tidy it with your hands.

Support replies

Answer tickets and messages without the typing strain.

Coding & product work

Dictate commit messages, issue reports, and review notes.

FAQ

Questions, answered

What is Secretary42?
Secretary42 is a local-first desktop dictation app. You press a global hotkey, speak, and your words are transcribed on your device with the bundled whisper.cpp engine and pasted where you’re working.
Is Secretary42 available now?
Yes. Secretary42 is generally available. You can buy a subscription and download it for macOS Apple Silicon and Windows x64.
How much does it cost?
It’s $9.99 per month or $99.99 per year. The yearly plan works out to about $8.33/month — nearly two months free compared with paying monthly.
Is my audio sent to the cloud?
No. Transcription runs on your device, and there is no cloud transcription fallback. Your audio and transcript text are processed locally.
What does connect to the internet?
Paddle checkout, license delivery, Whisper model downloads, app updates, support, and opt-in diagnostics may use online services. These are separate from the local dictation path.
Do I need an account?
No. Buy with your email through Paddle, receive a license token by email, and paste it into Secretary42 to activate the app. Dictation itself runs locally on your device once the app is installed and activated.
How does Paddle payment work?
Paddle is our payment provider and handles checkout and billing securely. You choose monthly or yearly billing and complete checkout through Paddle; the app itself does not process your card details.
What platforms are supported?
macOS on Apple Silicon and Windows x64.
Does it work offline?
Dictation works offline once you’ve downloaded a model and activated your license — transcription happens on your device. Checkout, license delivery, model downloads, and updates need an internet connection.
What diagnostics are collected?
Diagnostics are opt-in and off by default. If you enable them, only sanitized crash and performance metadata is shared — never your audio, transcript text, or local file paths.