thanks for looking! :) if anyone's interested, the voices use short (CHAR length.. actually usually around 20 samples..) delays as a waveguide vocal tract with kelly-lochbaum reflection junctions. you set how 'open' each segment is to make different sounds like drawing the cross-section. since this is 2 1/2 segments it can define about 1 1/2 formants to make vowels :)