macOS
Lulucat Speecho
Hold a key. Speak. Let go.
Dictation that ends where you were already typing
Lulucat Speecho transcribes speech while you hold the right Option key. The words appear in a panel at the bottom of the screen as you talk, and the moment you let go the finished text is pasted at the cursor — in a message box, a code editor, a search field, whatever had focus. Nothing has to be opened first and nothing has to be copied out afterwards: the sentence you just spoke is already in the document you were writing.
The whole interaction
One key, held down, does all of it.
Hold to talk
The right Option key starts a take when it goes down and ends it when it comes up. It works in every application, because the key is watched across the system rather than bound inside one window. The left Option key is left alone, so the special characters and shortcuts it types keep working as before.
Text while you speak
Audio is streamed for transcription while you are still talking, so words appear shortly after you say them rather than after you stop. Text the model has confirmed is shown in full contrast and text it may still revise is shown in grey, which tells you at a glance which part has settled.
Let go to paste
Releasing the key ends the stream, waits for the last words to be finalised, and pastes the result at the cursor. The text travels through the clipboard and the previous clipboard contents are put back a second later, so something you had copied earlier is still there when you need it.
Cheap to abandon
A press too short to contain speech is discarded, and so is a take during which another key was pressed — that is someone typing a special character, not dictating. Nothing is pasted in either case, so a mistaken press costs nothing and leaves nothing to undo.
Languages
Supports more than 60 languages, automatically recognizes the language being spoken, and handles multiple languages in the same sentence.
The panel while you talk
The panel appears at the bottom of whichever display the pointer is on, above full-screen applications and on every Space. It never takes focus, so the text field you were typing in stays active and the caret stays where you left it. A waveform on the left moves while audio is being sent and turns amber while the last words are being finalised, so a microphone that is picking nothing up is visible in the first second rather than after the fact.
Two system permissions
macOS asks for the microphone before any audio is recorded. Accessibility is what the other half of the feature needs: watching a modifier key from outside the app, and pasting into the app that has focus. Both are explained where they are requested, and the app keeps showing which one is missing until it is granted, including where to find it in System Settings.
What leaves the Mac
Audio is sent for transcription while you hold the key, and nothing else is uploaded. Recordings are not written to disk. The list of recent dictations is held in memory for the current run only and is gone when you quit. There is no account to create and no analytics.
Availability
Speecho is still being built and has not been released. It will require macOS 15 or later and will be distributed from this site rather than the Mac App Store: pasting text into another application is something the App Store sandbox does not allow, and that is the part of Speecho worth having.
Write to support@lulucat.ai if you would like to hear when it ships.