The following figures are analyses of sample outputs from the program. Three sound sources were given as separate inputs along with a specification for a small chamber orchestra. The output contains quarter tones and dynamic levels ranging from ppp to fff.
Although none of the results are recognizable as a car horn, boat whistle or alarm, the resemblances are noticeable when listening to the sources and orchestrations together. Regardless of how accurate they are, the chords are interesting constructs on their own, organized via an underlying complex of relationships based on the combination of the timbres of their individual notes.
The following figures show spectral FFT analyses of two sound recordings. One of them is of a boat whistle and the other is the orchestrated version, resynthesized by mixing together samples of instruments playing the notes in the chord.
The circles indicate a few frequency components that are shared between the two graphs. These common frequencies help preserve the pitch of the source while the similarity of their spectral envelopes shows a preservation of the sound's brightness and formant structure. The algorithm works, then, by matching both the perceived pitch and the distribution of energy in the source's spectrum.