Your browser already knows part of the delay between playing a sound and hearing it, and it reports that figure. Every result here is the delay on top of what the browser reports, measured from the moment the browser expects the sound to leave this device: the part a TV, soundbar or pair of headphones adds without telling anyone.
Flash and beep and Tap along compare sound with picture on this screen, so they show lip-sync error as you see it. Microphone times the sound alone: this device plays a short chirp and listens for it, so the reading also includes the time its own microphone takes, unless you calibrate first.
Sound is compressed, sent in packets and buffered in the earbuds. Low-latency codecs and game modes are often under 100 ms.
The microphone is only used while a measurement runs. What it hears is analysed on this device and never recorded or sent.