This is a solid-state Tesla coil. The primary runs at its resonant frequency in the 41 KHz range, and is modulated from the control unit in order to generate the tones you hear.
So just to explain a little further, yes, it is the actual high voltage sparks that are making the noise. Every cycle of the music is a burst of sparks at 41 KHz, triggered by digital circuitry at the end of a "long" piece of fiber optics.
What's not immediately obvious in this video is how loud this is. Many people were covering their ears, dogs were barking. In the sections where the crowd is cheering and the coils is starting and stopping, you can hear the the crowd is drowned out by the coil when it's firing.
This Tesla coil was built and is owned by Steve Ward. Steve is a EE student at U of I Urbana-Champaign. He and Jeff have been going to Teslathons, which is where they met.
It's been suggested that a good name for this coil would be the "Zeusaphone". "Thoremin" has also been mentioned, though personally I think we need Theramin type inputs for that.