The aim is a cross-browser, cross device, no-plugins, web-based talking chat avatar.
I looked at CrazyTalk, which seemed perfect, but sadly it turns out that that relies on the unity engine.
I then started to think about rolling my own by combining existing text to speech services and trying to pull phonemes out of an audio wave and make my own dictionary of phonemes to canvas shapes. That doesn't really seem to exist either (and even if it did, I'm not sure how I would work the timing on mouth movement to audio).
Its 2015, I feel like something like this should already exist and I shouldn't be trying to invent it.