Simple Whisper

A native macOS speech-to-text app powered by WhisperKit. Press a hotkey, speak, and the transcription is typed directly into any text field.

Requirements

• macOS 14.0+
• Apple Silicon (recommended) or Intel Mac
• Microphone permission
• Accessibility permission (for global hotkey and auto-type)

Build

xcodebuild build -scheme SimpleWhisper -configuration Debug \
  -project SimpleWhisper/SimpleWhisper.xcodeproj

Or open in Xcode:

open SimpleWhisper/SimpleWhisper.xcodeproj

No external dependency managers needed — WhisperKit is included via Swift Package Manager.

Usage

1. Launch the app — The Settings window opens. Go to the Model tab and download a Whisper model (Base is recommended for most users).
2. Grant permissions — Allow Microphone and Accessibility access when prompted.
3. Start transcribing — Hold Fn+Control (or your custom hotkey), speak, then release. The transcription is typed at your cursor.

Optional: AI Enhancement

Enable AI Enhance in settings to clean up transcriptions with an LLM. Requires an OpenAI or Anthropic API key.

Architecture

SimpleWhisper/
├── Models/          # AppState (Observable), enums, localization strings
├── Services/        # Audio recording, Whisper inference, LLM, hotkey, config
├── Views/           # SwiftUI settings, popover states, floating pill
└── Theme/           # Design tokens (colors, spacing, sizing)

Key design decisions:

• SwiftUI + AppKit hybrid — SwiftUI for views, NSPanel for the always-on-top floating pill.
• Observable singleton — AppState drives the entire UI via Swift Observation.
• Actor-isolated inference — WhisperService runs transcription off the main thread.
• No external assets — Sound effects are synthesized in-memory; icons use SF Symbols.
• Persistent config — Settings and history stored at ~/.simple-whisper/.

License

MIT