dictation // refined

Voice to cursor.
Type instantly.

An editorial, handcrafted dictation tool for macOS. Transcribe your voice in real time and paste it directly into your active window. Full offline security with zero network telemetry.

Download for macOS
100% on-device core ml models
Tick Key Model
active_app: terminal
[ hotkey: e.g. ⌥Space ]
"Hey team, just wanted to follow up on the design review. The new direction looks great—let's |"

Designed for focus.

Tactile workflows, raw hardware acceleration, and robust application formatting rules.

// 01_SPEED

Native Performance

Transcribe instantly. Custom Whisper execution optimized for Apple Silicon guarantees sub-100ms response times.

// 02_FLOW

Tactile Workflows

Designed as a background daemon. A single global key combination captures, formats, and types your thoughts with zero clicks.

// 03_CONTEXT

App-Aware Syntax

Automatically parses dictation style based on destination. Emits shell command logic, markdown notes, or conversational text depending on active application focus.

Dictation Input "git commit this is a good landing page"
Terminal Output git commit -m "this is a good landing page"
Slack Output "git commit this is a good landing page"
// 04_SECURITY

Absolute Isolation

Transcriptions are processed entirely in memory on your Mac. No network logs, no telemetry, and complete system privacy.

Mic Input
Core ML
Active App
Cloud

Zero clicks. Pure typing.

How Tick automates your spoken thoughts into the active document.

01

Trigger shortcut

Hold your configured global hotkey (e.g., + Space). Capturing begins instantly with a haptic chime.

02

Local Translation

The Core ML engine transcribes your voice directly in-memory using Whisper or Parakeet models.

03

Format Syntax

Tick analyzes the focused window. Shell commands, Slack chats, or notes are formatted on-the-fly.

04

Auto-Paste

The formatted text is typed directly into your active cursor position. The system plays a completion chime.

Tailor your speed and precision.

Choose from optimized on-device models. From featherweight models for instant command entry to heavy weights for precise long-form transcriptions.

100% Offline
Sub-100ms Latency
Tick Settings — Models
v0.7.1
Parakeet TDT v3 Default Multilingual • FluidAudio Core ML
Accuracy: ★★★★★ 610 MB
Whisper Medium English-only • WhisperKit Core ML
Accuracy: ★★★★☆ 1.5 GB
Whisper Large v3 Multilingual • WhisperKit Core ML
Accuracy: ★★★★★ 3.1 GB
64%

Speed at your fingertips.

Record / Dictate

Press and hold to start recording speech. Release to instantly transcribe and paste into the active window. (Customizable in settings)

+ Space (default)

Paste Last Transcript

Paste your last voice dictation into the focused input field again. (Customizable in settings)

+ + V (default)

Cancel Recording

Instantly cancel the current transcription process while recording.

Esc

Your data stays yours.

Tick is built to be a pure tool. We believe developer utilities should respect system isolation and secure input by default.

100% On-Device Engine

All audio recording, Whisper Core ML model loading, and speech processing happen locally on your Apple Silicon hardware. No data is ever uploaded to a server.

No Tracking or Telemetry

We do not log transcripts, track usage stats, or record analytics. There are no tracking scripts or third-party cookies anywhere in the app or on this site.

Frequently Asked Questions

How does App-Aware formatting operate?

Tick inspects the focused macOS application bundle identifier on trigger. If the active app is a terminal emulator or IDE (like VS Code or Terminal), it formats dictation into syntax-valid CLI options or code structures. If the app is a messenger (like Slack), it preserves natural conversation strings.

Are custom Whisper models supported?

Yes. You can import any custom Core ML-compiled Whisper model directly through the settings interface. The app scans your local directory and adds them to the available model list automatically.

Where are models saved on disk?

Whisper models are saved in standard app cache support paths: ~/Library/Application Support/com.dipxsy.Tick/models/. Parakeet models reside inside the sandbox container's FluidAudio paths.