Handy

Handy is a free, open source speech-to-text desktop application created by CJ Pais. Press a keyboard shortcut, speak, and Handy transcribes locally and pastes the text into the field where your cursor is. The latest release listed is v0.9.8, with builds for macOS, Windows, and Linux.

What It Is

Handy is a cross-platform dictation tool built around a single job: turning speech into text in whatever app you are typing in, such as a browser, editor, or chat app. Transcription runs on your own computer, and the project states that audio is not sent to the cloud. The app sits in the system tray or menu bar after setup.

How the Workflow Works

On first launch you choose a speech-to-text model to download. The docs describe Parakeet V3 for English and European languages, Whisper models for 99+ languages, and Cohere Transcribe for high-quality English on machines with a capable GPU. Silence is filtered with Silero voice activity detection. The default shortcut is Ctrl + Space, with push-to-talk or toggle modes, a separate shortcut for transcription with post-processing, and Escape to cancel. Shortcuts can be remapped in settings.

Customization and Control

Settings cover microphone selection, audio feedback, the recording overlay, silence detection, custom words, and AI post-processing to clean up or reformat text. Command-line flags can toggle transcription, toggle post-processing, or cancel, which helps on Wayland, where shortcuts must be set through the desktop environment. Custom Whisper GGML models can be placed in the models directory, and a community Raycast extension can control Handy.

Setup Notes

Linux builds ship as AppImage, .deb, and .rpm for x64 and ARM. The README also lists Homebrew and winget packages that are not maintained by the Handy developers. Linux text input works best with xdotool on X11 or wtype/dotool on Wayland. The source is MIT licensed, while the Handy name and branding are not open source.