The images in the video are all generated by a single piece of sculpture approximately the size of a telephone booth. The piece is made up of 16 separate 'visual systems', each with it's own small video camera. One camera is integrated into a diorama of a mechanical guitar player seated in a dimly lit grotto, one camera is pointed at the model of an exploding/imploding dome, one camera is mounted on a motor controlled panning device that collects images of the inside of the sculpture, and so forth. Each component in the piece (lights, solenoids, motors, etc.) is controlled by a corresponding MIDI note. A note might turn a single light bulb on and off. It might activate any number of devices that move.  Some notes trigger relays that provide power to the cameras and route video signals. Some notes control the application of rudimentary video effects.
MIDI is a language used to control electronic musical instruments. Because every element is controlled by midi information, it is easy to synchronise events to music using the same tools one would use to compose electronic music, in the same way you would control a musical instrument.  I use a PC to sequence the MIDI but you could, theoretically, just play it on a keyboard.

The sequence of a furry spirit drumming is expressed here in three different ways:







The last method of expression is the easiest to understand. The longest note, C#6 (where the numeral represents the octave), triggers a relay that supplies power to the relevant camera. B5,G#5, and C7 control the head and arms of the drummer. The grouping of notes on the bottom control a servo- the first grouping intializes a microprocessor (a BASIC Stamp that I programmed to control servos, which are motors whose positioning can be very accurately controlled) and the long note activates the servo, which moves very slowly in one direction as long as the note is held, then returns to its initial position when the note ends. The servo raises and lowers a golden log. This sequence has other components not seen here, such as lighting and video routing. I also designed light dimmers that could be controlled by MIDI controller data, which has 16,384 variation values. I found out later I could have bought them pretty cheaply instead.

Once a sequence has been carefully programmed, the entire video plays by itself with the press of a button.

Some images of the exterior of the piece can be viewed here.

P.S. TRAGICALLY, the guitar player was designed to not only strum and raise the neck of the guitar expressively, but also move his hand smoothly up and down the fretboard as well. The visual effectiveness of that scene was, I think, very powerful in large part due to this feature. It worked flawlessly for ages, but alas, it broke the day I shot the video. I can't fix it, because I gave little thought to serviceability at the time- the servo motor in question, which I bought from a Hong Kong dealer for 80 cents plus 3 dollars shipping, is deeply encased in the guitar (and a substance used for repairing boilers) and seems to be mortally seized.