Player not loading? Watch on YouTube
This tutorial covers Handy, which the speaker describes as a free, open source speech-to-text app for macOS, Windows and Linux. It focuses on dictating prompts and emails, with transcription running offline on the user's computer.
Model choice depends on language support and hardware. The speaker uses Canary 180M Flash, described as a 200 MB download, and says they would choose another model for Dutch. The setup covers separate left and right Command key shortcuts on Mac, push-to-talk or toggle recording, microphone selection, muting during recording and audio feedback.
Handy also has overlay settings and a direct paste option that the speaker uses without changing the clipboard. Transcription history lets users recover earlier text. Custom vocabulary helps with names and technical terms that speech recognition may mishear.
The optional experimental post-processing feature uses a separate hotkey and accepts custom prompts. The speaker says a local provider is available but demonstrates OpenAI with an API key and a model named GPT-5.6 Luna. That API setup is distinct from the app's local transcription. A cleanup prompt targets filler words and recognition errors; a rhyme demo takes longer and produces an unwanted result. The speaker checks history to compare the original transcript with the processed output, showing why generated edits need review.
