Handy: Open-Source Speech Recognition That Runs Entirely Locally Without the Cloud

Press a keyboard shortcut and speak, and the text is immediately pasted into the input field of the app you are currently working in. All processing happens on the user's computer, and audio data is not sent to external servers. It is a tool that allows you to use voice input safely in environments where privacy is a priority.

It supports Whisper models (Small, Medium, Turbo, Large) and Parakeet V3, a model optimized for CPU. If a GPU is available, it leverages acceleration, and automatically filters out unnecessary noise using Silero VAD. It works on all three platforms: Windows, macOS, and Linux, and is designed to be lightweight based on Tauri v2.

Unlike existing cloud-based speech recognition services, it does not require an internet connection. It is provided as open source (MIT license), allowing for customization, and can be extended with advanced features such as recording control and transcription history lookup through Raycast integration. It is designed as an accessibility tool that anyone can use for free.

It is suitable for work environments where data privacy is important, or for developers who need voice input even when offline. It can be used immediately by setting just one shortcut key without complex configuration, and is recommended for users who want to naturally add speech recognition to their existing keyboard input method.