VoiceInk

Dictate into any macOS app — press a hotkey, speak, and your words land where the cursor is.

VoiceInk is a native menu-bar dictation app. Hit Ctrl+Shift+S, talk, and when you stop it transcribes with Gemini and auto-pastes the text straight into whatever app has focus — no window to switch to, no copy-paste dance. I built it to scratch my own itch and now use it every day.

It's also my hands-on way to learn Swift / AppKit: a small, real, daily-driver app rather than a tutorial.

Status: early alpha — the core dictation loop is solid and in daily use. Next up: a model/engine selector, fully on-device transcription with local Whisper (no audio leaves your Mac), and auto-stop on silence.

Features

• Native AppKit menu bar app.
• Global Ctrl+Shift+S hotkey to start and stop recording.
• Microphone recording via AVFoundation.
• Gemini transcription via URLSession.
• Auto-copy and auto-paste into the focused text field.
• Menu bar SF Symbol icon with idle/recording state.
• Manual stop mode by default.
• Settings foundation for future auto-stop after silence.

Requirements

• macOS 13+
• Swift 6 toolchain / Xcode Command Line Tools
• Google AI Studio API key

Configuration

Create a local environment file:

cp .env.example ~/.voiceink.env

Then edit ~/.voiceink.env:

GOOGLE_API_KEY=your_google_api_key_here

You can also add the Gemini key from Settings... in the menu bar app. Keys saved from the app are stored in the macOS Keychain. The environment file is kept as a local fallback for development.

Build

Scripts/check.sh
swift run VoiceInkCoreSmokeTests
swift build --product VoiceInk
bash Scripts/build-app.sh

The app bundle is created at:

dist/VoiceInk.app

Run

open dist/VoiceInk.app

Install Locally

Scripts/install.sh

This installs VoiceInk to:

/Applications/VoiceInk.app

Using a stable /Applications path helps macOS keep Microphone and Accessibility permissions attached to the right app.

Quality

Scripts/format.sh
Scripts/check.sh

Scripts/check.sh runs swift-format in strict lint mode, the smoke test executable, and the app build.

Permissions

VoiceInk needs:

• Microphone to record audio.
• Accessibility to auto-paste the final transcript into the app that currently has focus.

If auto-paste does not work, open the menu bar icon and choose Request Paste Permission, then enable VoiceInk.app in System Settings.

Privacy

• API keys saved in the app are stored in the macOS Keychain.
• Recordings are written to a temporary WAV file and overwritten on the next recording.
• Transcription currently uses Gemini, so audio is sent to Google's API when you stop recording.
• VoiceInk does not include analytics or telemetry.

Roadmap

• Auto-stop after configurable silence.
• Engine/model selector:
  • Gemini API
  • OpenAI transcription models
  • Local Whisper
• Richer settings popover.

License

MIT. See LICENSE.