Push-to-talk and toggle recording need different gesture rules but should share one understandable capture state. A missed key release or repeated key event must not leave the app recording indefinitely or start overlapping sessions.
Define the state machine before tuning shortcut convenience. Users need to know when capture starts, when it stops, and whether text is still processing after the microphone is inactive.
Specify gestures and state transitions
For hold mode, state which press starts capture and which release ends it. For toggle mode, state which action starts and stops, how repeated events are handled, and what Escape or another cancel action means.
Keep idle, starting, listening, stopping, recognising, review, inserting, cancelled, and failed states distinct where the workflow needs them. A single recording boolean cannot explain all those stages.
Exercise a fictional editor session
Use harmless text in a controlled editor. Test a short hold, long hold, rapid tap, repeated keydown, modifier combinations, focus change, permission denial, device loss, and app sleep while a gesture is active.
For toggle mode, test repeated activation while recognition is finishing and while a preview is open. Define whether it starts a new session, stops current work, or is ignored with clear feedback. Do not let two requests compete to insert into the same destination accidentally.
No shortcut compatibility or timing result is claimed here. Record exact hardware, macOS version, input device, and target applications before describing observed behaviour.
Make cancellation and visibility consistent
Show actual microphone activity and a separate processing state. Stopping capture is not the same as cancelling the pending transcript or insertion. A cancellation should invalidate the relevant session so late output cannot revive it.
Provide an accessible way to discover and change the gesture where supported. A global shortcut can conflict with system or application behaviour, so a default needs a tested fallback and clear limitations.
The tradeoff is speed versus accidental activation and state complexity. Hold mode can offer a physical boundary while toggle mode supports longer speech without sustained key pressure. Neither is universally better; evaluate the user's task and input setup.
The Mac dictation workflow guide explains asynchronous capture and insertion. StackSpeech's portfolio description documents hold-or-toggle capture and a live listening indicator. This gesture matrix is proposed guidance, not a universal reliability claim.
Questions about recording gestures
Does releasing the key mean all text is final?
No. Capture can stop while recognition still finalises. Show processing separately and preserve the session boundary.
Should repeated key events create new sessions?
Define a safe rule for the chosen shortcut mechanism. Uncontrolled repeated activation can create overlapping capture or insertion requests.
What should a review test first?
Start, interrupt, cancel, and restart using both modes while focus changes. Bring unclear states to a Mac input-control review.