simpletranscribe

A minimal macOS menu bar app that transcribes system audio in real time using Apple’s Speech framework (SFSpeechRecognizer) and ScreenCaptureKit for audio capture. The app provides a simple window to view the live transcript, a Start/Stop control, an “Always on Top” toggle, and menu bar access.

Features

• Real-time transcription of system audio (not just microphone)
• Menu bar control to Start/Stop and bring window to front
• “Always on Top” toggle to keep the transcript window floating
• Partial results appear continuously; finalized segments are indicated in the status

Requirements

• macOS 14.0+ (Sonoma)
• Xcode 15+ (recommended)
• App requires the following permissions:
  • Screen Recording (to capture system audio via ScreenCaptureKit)
  • Speech Recognition (for SFSpeechRecognizer)

Note: Some locales require a network connection for recognition unless on-device models are available. This app defaults to en-US.

Getting Started

Open and run in Xcode

1. Open simpletranscribe.xcodeproj in Xcode.
2. Select the simpletranscribe scheme and a “My Mac” destination.
3. In Signing & Capabilities, select your Development Team if prompted.
4. Build and run (⌘R).
5. On first run, macOS will prompt for permissions:
   • Allow Screen Recording for the app (System Settings → Privacy & Security → Screen Recording).
   • Allow Speech Recognition (System Settings → Privacy & Security → Speech Recognition).

Using the app

• Click the menu bar item (waveform icon) to Start/Stop transcription, toggle “Always on Top,” or show the main window.
• Press Space to Start/Stop from the main window.
• The transcript appears in the scrollable text area.
• Status messages (e.g., “Listening…”, errors, or “Finalized segment”) appear in the UI and the menu bar menu.

Configuration

• Locale: The recognizer defaults to en-US. You can change the locale by editing TranscriptionService initialization in ContentViewModel.
• On-device recognition: The code currently sets requiresOnDeviceRecognition = false. Change this if you want to force on-device only where available.

How it works (high level)

• AudioCaptureManager uses ScreenCaptureKit to start a stream that captures system audio (excluding the app’s own audio) and forwards CMSampleBuffers.
• TranscriptionService wraps SFSpeechRecognizer, feeding it audio buffers and publishing partial/final text updates.
• ContentViewModel wires capture → transcription, exposes state to SwiftUI, and handles Start/Stop.
• ContentView and MenuBarExtra provide the minimal UI.

Troubleshooting

• No audio or text is appearing:
  • Ensure you granted Screen Recording and Speech Recognition permissions.
  • Quit and relaunch the app after changing permissions in System Settings.
• “Speech recognition not authorized”:
  • System Settings → Privacy & Security → Speech Recognition → enable for this app.
• Build/Signing errors:
  • Open the project in Xcode, go to the simpletranscribe target → Signing & Capabilities, and set your Development Team.
• Locale issues or low accuracy:
  • Try a different locale identifier (e.g., en-GB, fr-FR) in the TranscriptionService initializer.

Development Notes

• This project uses SwiftUI and Apple frameworks only (no external dependencies).
• Item.swift includes a placeholder SwiftData model that isn’t currently used by the app logic.
• ScreenCaptureKit requires relatively recent macOS; this project targets macOS 14+.

Roadmap (ideas)

• Export/save transcripts
• Microphone-only mode option
• Choose capture source (specific app/window)
• Configurable locale and on-device-only toggle in UI

License

Add a license of your choice (e.g., MIT, Apache 2.0) before publishing.