Reacts to the beat, tries to automatically compensate for continuously low or high amplitudes.
This originally had to analyze speech, but ended up to be just goofing around with Eigen/FFT.
Currently has no support for stereo tracks. They do work, but produce incorrect results.