Kaze

Hold a global hotkey, speak, and the transcribed text is automatically pasted into whatever app you're using.

Kaze.mp4

How it works

1. Hold Option + Command -- recording starts and a floating waveform overlay appears at the bottom of your screen
2. Speak -- your audio is captured and transcribed
3. Release -- transcribed text is pasted into the focused app (your clipboard is preserved and restored)

The app lives entirely in the menu bar with no Dock icon. The only UI during use is a minimal floating pill showing audio levels and live transcription.

Features

• Dual transcription engines -- choose between Apple Dictation (SFSpeechRecognizer, zero setup) or on-device Whisper via WhisperKit
• Apple Intelligence enhancement -- optionally post-process transcriptions with on-device Foundation Models to fix grammar, punctuation, and formatting (macOS 26.0+)
• Global push-to-talk hotkey -- works from any app without switching focus
• Animated waveform overlay -- floating non-activating panel with real-time audio level bars and scrolling transcription text
• Clipboard-safe -- saves and restores your clipboard contents around each paste

Tech stack

• SwiftUI + AppKit -- SwiftUI for the settings and overlay views, AppKit for the menu bar, floating panel, clipboard, and simulated key events
• Speech framework -- SFSpeechRecognizer for real-time streaming dictation
• WhisperKit -- local on-device OpenAI Whisper model for offline transcription
• Foundation Models -- Apple Intelligence on-device LLM for text enhancement
• CGEvent -- global hotkey detection via a low-level event tap
• Combine -- reactive state bridging between transcription engines and the UI

Requirements

• macOS 26.0+
• Xcode 26+
• Accessibility permission (for global hotkey)
• Microphone permission

Building

Open Kaze.xcodeproj in Xcode and build. WhisperKit is the only external dependency and is resolved automatically via Swift Package Manager.

License

MIT