I made this voice box so I could answer questions in Italian for the Rome Maker Faire, now I only hope I understand the question. :~
Basically it is a wave shield like circuit and a bi-colour square pixel 8x8 matrix. It took a lot of decoupling to keep the noise from the matrix multiplexer out of the audio. There are a sequence of 10 faces with varying degrees of open mouth, every 20mS a sample of the audio is taken through an envelope follower and the reading mapped to one of the 10 faces. So basically the louder the audio is the more open is the mouth. So given that, the effect works remarkably well.
Thanks to the guys in the Italian section for the translations and audio files.