Press listen and hold your device near the music. Earshot needs about 3 to 6 seconds.
No music nearby? That button plays a random song from the library through your speakers from a random point, so you can test the mic.
Nothing yet.
The audio is cut into 93 ms slices and each one is split into frequencies with an FFT.
Only the loudest point in each frequency band survives. Noise and volume barely change where those peaks are.
Every peak is paired with a few that come right after it. Two frequencies and the gap between them make one fingerprint.
The right song is the one where dozens of fingerprints agree on the same start time. That's the straight line in the chart.
A few seconds of a song from the library, recorded or cut from a file. MP3, WAV, M4A or OGG.
Drop full songs here and Earshot fingerprints them on your device. Nothing is uploaded. Afterwards it can name them from a few seconds of mic audio.