Player not loading? Watch on YouTube
This tutorial walks through Handy, a free, open source dictation app with downloads for Mac, Windows and Linux. The speaker uses the Linux AppImage and demonstrates inserting dictated text into another application with a hotkey. Setup covers microphone selection and the choice between push-to-talk and a shortcut that starts and stops recording.
The speaker says transcription runs on-device. Users download and activate a speech model, then can switch between installed models. The tutorial recommends testing models against the machine's resources, accent and language. Advanced settings include a live transcription overlay and a model unloading policy: keeping the model in memory avoids repeated loading, while unloading can suit machines with fewer resources.
Output options include typing characters directly, clipboard insertion and an external script. The speaker also shows custom vocabulary, but reports that an overly broad word list caused unwanted substitutions.
Optional post-processing sends the transcript and a cleanup prompt to an LLM before inserting the result. The demonstration uses Gemma 4 through a target on the speaker's local network; Ollama and remote OpenRouter are discussed as options. Local processing requires hardware capable of running the model. Separate shortcuts select raw transcription or cleanup for spelling, punctuation and paragraph breaks. The final example uses dictation to instruct an AI agent.
