Claps trigger it. Talking, music, and noise don't. All sound is synthesized live.
How it works: A clap is a sharp broadband transient — a sudden energy spike that's bright (lots of high frequency) and noisy (flat spectrum). Speech is low & tonal; bass is low; whistles are tonal. All three checks must pass at once, so only real claps count. Tune the sliders if your room is loud.