It's not much more than taking a spectrogram of the speech passage (could be any sound really), quantising it to equal temperament scale and converting intensities to velocities and then into MIDI data. But...But...BUT...it's cool :D Cheers Chris