Quill Join the waitlist

Speak, and it's typed,
nothing leaves your Mac.

AI that turns speech into clear, polished writing in every app. No subscription required.

Free, for macOS on Apple Silicon.

Works in every app you already use
Gmail Cursor Claude ChatGPT LinkedIn Slack WhatsApp Instagram Facebook Reddit SMS
Dictate an email Chat on Slack Drop a quick note Write code comments
Stored locally

Dictation

Offline Engine Ready
TODAY
09:14 AM Alright, let's get the new onboarding steps documented. Hold the fn key, speak, and release.
10:30 AM I need to drop a quick code comment here to explain why we're using a plain HStack instead of NavigationSplitView.
11:45 AM Transcribing on-device is so much faster. Under two hundred milliseconds and it types right into Slack.

Insights

Dictation Frequency
M
T
W
T
F
Quiet Cleanup Rate 94.2%
Top Focused App VS Code

Style Profiles

Medium Cleanup rewriting

Opt-in

Uses API keys locally to adjust speaking tones to match specific context patterns.

Default (Light)
Professional
Casual

API Provider Configuration

Provide your own key. Quill never uploads this key or your prompt parameters anywhere.

At a glance
2,840 total words
145 words / min
12 day streak
Local only

Dictation without the ceremony

1

Hold fn, speak, release

No record button, no stop button. Quill records audio only while the hotkey is depressed, eliminating silent tail ends.

2

Types where your cursor is

No copy-paste, no app switching. Dictated text streams straight into your active text editor or input field as keyboard strokes.

3

Nothing leaves your Mac

Transcription runs locally on the macOS Neural Engine. No cloud servers are contacted, ensuring complete privacy.

On-device, always

Runs on the Neural Engine

Every word is transcribed locally via CoreML. Nothing is uploaded, recorded remotely, or used to train models. There's no server in the loop at all.

🧠
Auto Cleanup

Filler words gone before you see them

Light mode strips "um," "uh," and stray filler words locally, instantly, for free.

Medium goes further with tone-aware rewriting, using your own API key, which is entirely opt-in.

"So um I think we should uh push the build now."
"I think we should push the build now."
Switch models on the fly

WhisperKit or Parakeet, your call

Swap transcription engines live from the menu bar with no app restart needed.

Quill handles downloads in the background, displaying real-time progress the first time a model's needed.

WhisperKit Base (English Only)
Active
Parakeet 110M (Fast)
Not Downloaded
Works everywhere you type

Even the apps that fight back

Native text fields, browser forms, and Electron apps like Slack and VS Code: Quill's injection engine handles all of them, including the ones that normally drop synthetic keystrokes.

See your own patterns

Dictation history & Insights

Track words-per-minute, speaking habits, streak calendars, and a local record of cleaned up phrases.

All stats are computed locally from text logs on your system and are never synced anywhere.

Streak 12 Days 🔥
Words Dictated 2,840

Free. Not "free trial."

SaaS pricing models don't belong in a local app.

$0 forever

On-device transcription and Light Auto Cleanup cost nothing and always will. If you turn on Medium's tone rewriting, you bring your own OpenAI or Anthropic key and pay that provider directly. Quill never sees a cent either way.

Join the Waitlist

Wispr Flow and Gladio charge a monthly subscription for the same core idea. Quill doesn't.

Actually private, not just promised.

On-device transcription

Every word is processed on your Mac's Neural Engine. Audio data is kept in-memory during processing and immediately discarded.

No account, ever

There's no sign-up, no email verification to log in, and no central user database to be breached. You own your copy of the app.

No telemetry

Quill doesn't track what you say, how often you dictate, or what applications you type into. We never track usage data.

One clearly-marked exception

The optional Medium tier sends text snippets to your chosen LLM provider (OpenAI or Anthropic) via your own API key. This is opt-in only.

Frequently Asked Questions

A free, local-first dictation app for macOS. Hold a shortcut hotkey, talk, and releasing the key automatically types whatever you spoke into whichever text container has current cursor focus.
No. Core transcription runs 100% offline using WhisperKit/Parakeet models on the Apple Neural Engine. The only feature requiring internet is the optional Medium cleanup tier which rewrites transcripts using external LLMs via your own API credentials.
Audio is captured directly in volatile RAM, processed instantly by the local neural model, and erased immediately. Nothing is cached on disk, saved locally, or transmitted to any server.
Yes. Running transcription on your own machine costs us nothing, so it costs you nothing. Light Auto Cleanup is also completely free. If you use Medium's LLM tone-cleanup rewriting, you are billed directly by your chosen provider (OpenAI or Anthropic) for key usage; Quill has no markup.
Quill is optimized for Apple Silicon Macs (M1, M2, M3, M4 series or newer) running macOS 14 (Sonoma) or newer to leverage CoreML model accelerations on the Apple Neural Engine.
Wispr Flow and Gladio utilize cloud servers to process audio and require paid monthly or annual subscriptions. Quill is local-first, runs offline on your own GPU/Neural Engine, records no telemetry, and is completely free.

Ready to stop typing?

Join the waitlist. Free for macOS on Apple Silicon.

Free, forever 10-second setup Nothing leaves your Mac