This is the second test and example of an audio visualization done in Blender. The video is a result of a python add-on for blender in development. Several fourier transforms were done automatically (104) on a logarithmic frequency scale. The speakers are shape-key driven with a driver tied to a broad base FFT. The mesh is an automation process that creates the mesh from data in the full range FFT. No additional modules or add-ons were used, only stock blender 2.61.
The video was rendered using OpenGL on an Nvidia 8500 GT instead of the traditional raytracers. Custom vertex and fragment shaders were used along with one point lamp. The scene totals 743537 faces (quads).