There is software that does just this. If you have access to a Mac, a program named MetaSynth uses 2D pictures and Fourier transforms to synthesise audio. I suspect that RDJ used this.
http://faculty.washington.edu/dillon/PhonResources/javoice/vowjavoice2.html .
Javascript utility for converting images into sound. Developed in researching sight for the blind.
Damn, I was gonna try this, but RDJ beat me too it by the looks of it. I just had to work out a way to draw pictures using whitenoise and filters...