This is the first product of a series of experiments into visualizing music via FFT and Perlin noise. This is what it looked like after 1 day of development.
There is very little scrubbing performed on the FFT output, so this visualizer is not as consistent as the later version. However, because it lacks adaptive scrubbing it also produces very interesting artifacts at some points, such as about four minutes in, where there is a loud rise that results in a constellation of attractors which the later version would have scrubbed out.
I also matched the shutter speed of the camera to the intensity of the music by adding motion blur whenever the track is quiet.