I found that it is difficult to control the data for the 3D graphics to be generated. So, I attempted to get rid of the function that I didn't want to work with, and made it as simple as possible.
Max patch : File:hg20181114sirin.maxpat
The position of the spheres is the only thing that is being controlled by the audio input, using different scales for each sphere. So, it is just one audio input creating diverse results.