Guides for Mac and Windows

If the idea is new, read What is on-device transcription, then How local transcription works. If you already know the job — video, speaker labels, a recording you already have — go straight to the group that matches: Calls and meetings, Dictation, or Files and video. The measured pages are for choosing a model and seeing where models fail.

Guides explain how voice becomes text, one idea or one job per page. Compare puts Lesskeys next to a named tool — Superwhisper, Otter, Rev — point by point. Industries covers what one line of work needs: lawyers, therapists, journalists. Three questions, one product.

Yes. We ran them on a Mac with Apple Silicon against public recordings and checked the output against human transcripts, rather than repeating vendor claims. We publish the inconvenient results too, including a 53-minute recording that came back 3% transcribed while the app reported it had finished.

The built-in microphone on any Apple Silicon Mac is enough. A studio microphone buys very little accuracy — a quiet room helps more. Bluetooth headsets are the weak point: in headset mode the microphone quality drops noticeably.

No. Recording starts only when you press the dictation hotkey or when you start a transcription job. On a Mac the speech model unloads from memory after an idle period you set. Transcribing a long file works your Mac hard while it runs, so expect fan noise and battery use.

That is text to speech, the opposite direction, and Lesskeys does not do it. macOS includes it: open System Settings, then Accessibility, then Spoken Content, and switch on Speak Selection. Select text and press the shortcut shown there. Lesskeys only turns speech into text.