ezspeak

Voice-to-text transcription using OpenAI Whisper API. Press a button, speak, and text appears.

Setup

npm install
npm start

Enter your OpenAI API key when prompted.

Usage

Press button → Speak → Press again → Text pastes automatically OR Hold button → Speak → Stop holding button → Text pastes automatically

Build

npm run build

Project Structure

• Root
  • main.js: Electron main process entry point.
  • preload.js: Bridge between Electron's main process and renderer.
  • package.json: App configuration and builder settings.
• src/ (Backend logic and services)
  • audioRecorder.js: Handles microphone capture and temp file management.
  • transcription.js: Routes transcription between OpenAI API and local Whisper.
  • localTranscription.js: Wrapper for local whisper-bin execution.
  • hotkeyManager.js: Global key interception and smart hold/toggle logic.
  • clipboardManager.js: Handles auto-pasting functionality.
  • config.js: Persistent settings management (electron-store).
  • logger.js: Customized file and console logging.
• renderer/ (Frontend UI)
  • settings.html/js/css: Main configuration interface.
  • recording-indicator.html: Floating UI overlay shown during active recording.
  • microphone-selector.js: Logic for detecting and selecting audio inputs.
• website/: Next.js landing page (Splash page) hosted on Vercel.
• whisper-bin/: Pre-built Whisper.cpp binaries for local transcription.
• scripts/: Utility scripts like download-model.js for local Whisper models.
• assets/: Store app icons (.ico, .png).