OpenVoice: Local-First Voice Typing for macOS

Introduction

OpenVoice is a privacy-focused voice typing application designed specifically for macOS users. It allows users to dictate text by holding a shortcut key, speaking, and having the text appear at their cursor or be copied to the clipboard. Unlike cloud-based dictation services, OpenVoice ensures that audio never leaves the user's machine. This design prioritizes data privacy and allows the app to work fully offline without an internet connection.

Benefits

OpenVoice offers several key advantages for users who value security and performance. First, it uses a privacy-first architecture where all audio processing happens locally on the user's device. No audio data is sent to external servers. Second, the app is fully functional without an internet connection, making it reliable in areas with poor connectivity. Third, it starts with a lightweight model that is only 75 MB in size to ensure fast performance. Heavier models are hidden by default and only downloaded when the user explicitly requests them. Finally, the app uses a warm model cache, meaning the active transcription model stays in memory after the first run. This significantly speeds up subsequent transcription tasks.

Use Cases

OpenVoice is ideal for anyone who needs to type quickly while maintaining strict control over their data. It works well for writers, journalists, and students who need to draft documents without worrying about their voice data being stored in the cloud. The app operates as a menu bar accessory, so it stays out of the way in the Dock. A floating display appears only when recording or transcribing. Users can choose between a push-to-talk mode or a toggle recording mode to suit their workflow. The app guides users through necessary permissions and model downloads before use, ensuring a smooth setup process.

Pricing

Pricing details are not available for OpenVoice as it is currently a private project in development.

Vibes

Public reception and reviews are not available for OpenVoice as it is currently a private project in development.

Additional Information

OpenVoice is currently a private project that is in progress. It is built using a modern stack that includes Tauri 2, which combines Rust and Web technologies. The frontend is built with React and TypeScript. For audio processing, it uses a Rust implementation of Whisper that can utilize Apple's CoreML for on-device acceleration. The project requires Rust, Node.js, and Xcode Command Line Tools for development. Models and application settings are stored in the user's library directory at ~/Library/Application Support/com.openvoice.app/.