Player not loading? Watch on YouTube
Handy is an open source dictation app that the presenter describes as free, offline and MIT licensed, with no account or subscription. The walkthrough explains how a keyboard shortcut starts recording through either a toggle or push-to-talk mode. After release, Handy transcribes the audio with the selected model and pastes the text into the focused input. Voice activity detection filters out silence.
The technical overview covers Tauri, a Rust backend and a React and TypeScript settings interface. Model choices include Whisper and Parakeet, with speed and accuracy indicators in the app. The presenter uses Parakeet Unified, a 0.6 billion parameter model, on an M2 Max MacBook. He cites the model's reported speed of roughly five times real time on a mid-range i5 CPU. He also explains how to add a fine-tuned Whisper GGML model to the models folder and restart the app.
The comparison uses a spoken passage containing SQLite, COBOL and AI product names. In the presenter's assessment, Handy handles punctuation and technical terms better than Google Docs voice typing. Whispr Flow produces the best result when he repeats the passage as closely as possible. He still prefers Handy for its cost and local processing.
