Is the lipsync behavior armed when you make the recording? Can you record the microphone in CH, play it back (and hear it), then use compute lipsync from audio ? does that work? Finally, in the lipsync and face behaviors, there are view/handle sections you can expand - is it being bound through to the mouth visemes correctly? If an older puppet, i think there were some changes in how it found layers - I recall someone else having to do minor adjustments for the visemes to be located correctly for them as well.