A macOS menu bar app that records your microphone and transcribes speech to text using Whisper via WhisperKit. The transcription is automatically copied to your clipboard and pasted at the cursor.

Built as a learning project to understand how apps like MacWhisper work under the hood.

• Xcode 15+
• macOS 13+

Open VoiceApp/VoiceApp.xcodeproj in Xcode and press Cmd+R.

1. A mic icon appears in the menu bar (top-right of screen)
2. Press and hold Fn (globe key) to start recording — or click the menu bar icon to open the popover and use the button there
3. A floating mic circle appears on screen while recording
4. Release Fn (or click Stop & Transcribe) to transcribe
5. The transcription is copied to your clipboard and auto-pasted at the cursor
6. View past transcriptions in the History tab of the main window

Grant microphone and accessibility permissions on first run.

VoiceAppApp (@main)
├── AppDelegate
│   ├── NSStatusItem + NSPopover → RecorderView (SwiftUI popover)
│   ├── FloatingMicView          — overlay circle shown during recording
│   └── HotkeyManager            — global Fn key monitor (requires Accessibility)
└── Window scene → TabView
    ├── HistoryView   — shows past transcriptions
    └── SettingsView  — model selection, download, delete

RecordingController (@Observable)  ← shared by AppDelegate + RecorderView
├── AudioRecorder        — AVAudioEngine tap → AVAudioConverter → 16kHz mono Float32
├── TranscriptionService — WhisperKit → NSPasteboard + CGEvent auto-paste
└── TranscriptionHistory — in-memory list of transcription records

State machine (AppState): idle → recording → transcribing → done(text) | error(message)

Open VoiceApp → Settings (Cmd+,) to choose the WhisperKit model; it downloads automatically on first use. Models are cached at ~/Library/Application Support/VoiceApp/WhisperKitModels/ and can be deleted from the Settings UI to free disk space.

• Support customising keyboard shortcuts (Cmd+Z)
• Clicking the floating should show settings
• Closing the settings shouldn't stop the app, it should still keep the floating mic
• On start up, load the model first, show a spinning circle
• Allow continuous speech without lifting the key. Automatically detect pauses to transition into typing.
• Onloading, the mic icon shouldn't show up
• Allow custom dictionary (and post-processing)