4 ms·
They buffer the audio in a ring buffer constantly, so that the audio can be checked for a wake word. If the wake woRd triggers, then they send the contents o
by hiddencost 11y ago
They buffer the audio in a ring buffer constantly, so that the audio can be checked for a wake word. If the wake woRd triggers, then they send the contents of the buffer. The buffer shouldn't be super big. (E.g., one or two seconds). They use it to train their wake word model.