Audio Control

A macOS desktop application for real-time speech-to-text transcription using local processing. Built with Svelte frontend and Tauri (Rust) backend, this application provides minimal latency transcription using the Whisper.rs engine.

Features

• Real-time speech-to-text transcription
• Local processing - no data leaves your computer
• Minimal latency
• macOS native desktop application
• Built with Svelte and Tauri for optimal performance

Prerequisites

• macOS
• Node.js (v16 or later)
• Rust (latest stable version)
• Whisper model files (see below)

Whisper Models

This application uses Whisper.rs for local speech recognition. You'll need to download the appropriate model files:

1. Visit Whisper.rs GitHub for model information
2. Download the model files (recommended: ggml-base.en.bin for English)
3. Place the model file in the application's model directory

Development

To run the application in development mode:

# Install dependencies
npm install

# Start the development server
npm run tauri dev

Building

To create a production build:

# Build the application
npm run tauri build

The built application will be available in the src-tauri/target/release directory.

Usage

1. Launch the application
2. Select your audio input device
3. Start speaking - your words will be transcribed in real-time
4. The transcribed text will appear in the application window

Notes

• For best performance, use a model size appropriate for your hardware
• The application processes all audio locally - no internet connection required
• Initial transcription might take a moment as the model loads into memory

License

[Your License Here]