Dictation

Set a dictation key in Settings › Hotkeys. Hold it, speak, let go. A small HUD shows the words as they are recognised and the final text is pasted where your cursor was.

Recognition uses macOS 26’s SpeechAnalyzer, which runs on the Mac. Audio never leaves it. This needs macOS 26 and the microphone permission, requested on first use.

There is a palette row too, Dictate, and cleat://dictate toggles it from a script or a Stream Deck.

What it needs

Dictation is a macOS 26 feature. Cleat uses SpeechAnalyzer, the on-device recogniser Apple shipped that year, so the audio is processed on the Mac and never uploaded. The first use asks for the microphone permission; macOS handles that prompt, and you can withdraw it later in System Settings under Privacy and Security.

If you are on macOS 15, the Dictate row still appears in the palette but explains that it needs a newer system rather than failing silently.

Where the text goes

Cleat pastes the recognised text into the app you were in when you held the key, at the cursor. That path uses the same Accessibility permission as clipboard paste back, so if window snapping already works, dictation will too.

Punctuation follows what you say: “comma”, “full stop” and “new line” are recognised as commands rather than words, because that is SpeechAnalyzer’s behaviour and Cleat does not modify the transcript.

Comparing it to the system dictation key

macOS has its own dictation, bound by default to a double press of a modifier. The difference is where the words land. System dictation types into whatever has focus and needs that field to accept a text cursor. Cleat pastes, which works in a few places system dictation struggles with, and it stops the moment you let go of the key rather than waiting for a pause.