WhisperKey

Privacy-focused local dictation for macOS

🚀 v1.0.2 Released! (July 18, 2025): Fixed keyboard focus issues and improved installer experience! See release notes

WhisperKey brings the power of OpenAI's Whisper AI to your Mac for fast, accurate speech-to-text that works in any app. Your voice never leaves your device.

Features

• 🎙️ Works Everywhere - Dictate into any text field in any app
• 🔒 100% Private - All processing happens locally on your Mac
• ⚡ Fast & Accurate - Powered by Whisper AI with Metal acceleration
• 🎯 Simple to Use - Just tap your hotkey and speak
• 🎨 Native Mac App - Clean menu bar interface that feels right at home
• ⏱️ Recording Timer - See how long you've been recording with live audio levels
• 🔊 Smart Audio Feedback - Different sounds for success vs clipboard saves
• 📋 Clipboard Fallback - Automatically saves to clipboard when not in a text field
• ⚙️ Customizable - Adjust silence detection, microphone sensitivity, and more

Quick Start

1. Download & Launch - Get the latest release and run WhisperKey
2. Complete Setup - Follow the onboarding wizard
3. Grant Permissions - Allow microphone and accessibility access
4. Start Dictating - Tap Right Option (⌥) to start/stop recording

Note: As of v1.0.1, whisper.cpp is bundled with the app - no separate installation needed!

Default Hotkey

Right Option (⌥) - Tap once to start recording, tap again to stop

You can change this in Settings to F13 if preferred.

Installation

Option 1: Download Release (Recommended)

1. Download the latest WhisperKey.dmg from Releases
2. Open the DMG and drag WhisperKey to Applications
3. Important - Security Steps:
   • Right-click WhisperKey in Applications and select "Open"
   • You'll see a warning about an unidentified developer
   • Click "Open" to proceed
   • If the above doesn't work:
     • Open System Settings → Privacy & Security
     • Scroll to the bottom
     • Look for "WhisperKey was blocked..." and click "Open Anyway"
4. Follow the onboarding wizard which will:
   • Guide you through permission setup
   • Download AI models automatically
   • Configure your preferences

Option 2: Build from Source

# First, install whisper.cpp
git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp
WHISPER_METAL=1 make -j

# Then build WhisperKey
git clone https://github.com/BSPLAZA/WhisperKey.git
cd WhisperKey
swift build

Requirements

• macOS 12.0 or later
• Apple Silicon or Intel Mac
• ~500MB disk space for AI models
• whisper.cpp (v1.0.1+ includes bundled binary or guides installation)

Tips

• Speak clearly and at a normal pace
• WhisperKey automatically stops after 2.5 seconds of silence
• The menu bar icon turns red while recording
• A floating window shows recording time and audio levels
• Press ESC to cancel recording
• Cannot dictate into password fields for security
• Success message shows word count inserted

Current Limitations

v1.0.1 Fixed:

• ✅ Now includes bundled whisper.cpp binary
• ✅ Models download to ~/.whisperkey/models/ (no manual setup)
• ✅ Improved error handling and disk space checks
• ✅ Better model detection across multiple paths

Still Present:

• App is unsigned - right-click and "Open" on first launch
• English models only (multilingual coming soon)

Coming in v1.1:

• Code signing and notarization (no more security warnings)
• Multilingual support (UI and models)
• Custom vocabulary support

Models

WhisperKey includes three AI models:

• Base - Fastest, good for quick notes
• Small - Balanced speed and accuracy (default)
• Medium - Best accuracy, slower

Privacy

WhisperKey is designed with privacy first:

• No internet connection required
• No data leaves your Mac
• No analytics or tracking
• Open source

Building from Source

Prerequisites:

• Xcode 14.0 or later (or Xcode Command Line Tools)
• whisper.cpp built with Metal support

# Clone and build
git clone https://github.com/BSPLAZA/WhisperKey.git
cd WhisperKey
swift build

# Or open in Xcode
open WhisperKey/WhisperKey.xcodeproj

Contributing

We welcome contributions! Please see our Contributing Guide for details.

Known Issues

• Brave Browser: Requires manual click after dictation (press Space then Enter in URL bar). This is a Brave security feature.
• System sounds may be transcribed (e.g., notifications appear as "bell dings")
• No formatting preservation (plain text only)
• Single audio device support (doesn't handle switching mid-recording)
• Custom vocabulary not yet supported

Support

Having issues? Check the troubleshooting guide or open an issue.

License

WhisperKey is open source software licensed under the MIT License.

---

Made with 🎤 for the Mac community