Voice Assistant Settings Block for React

It counts runs of vowels in each word, less a silent final e, which is a fair guess for English. Fewer than two syllables is too short and would wake by mistake, two works, three or four sounds distinct, and more is long. It is a hint, not a recogniser: in your app test a phrase against your wake engine's own false-accept rate.

A 30-word reply takes 12 seconds at normal speed, so the time is 12 divided by the speed, rounded to whole seconds: about 11 s at 1.10 times and about 8 s at 1.5. The slider moves in steps of 0.05 and reads its value as words. Change REPLY_SECONDS to match your voice's pace.

It opens an alert dialog that says how many requests will be deleted and that it cannot be undone, with Keep history as the safe choice. Confirming sets the typed count to zero and says "Deleted 142 requests in this demo." Deleting happens at once and is not part of Save, so Discard does not bring the history back. With nothing kept the button stays but says there is nothing to delete.

Everything. Nothing is listening, no sound is made, Save keeps the settings in this page's state and the history count is typed. In your app save the settings with the device, speak with your own text-to-speech engine at the chosen voice and speed, and delete history from the server before you tell anyone it is gone.

Every control is named, the voice cards are a radio group you move through with the arrow keys, and the sample reply carries the language of its text so a screen reader switches voice. The status line is a polite live region. On a phone the rows stack, the controls take the full width and the action foot is pinned to the bottom of the screen.