I'll not get into all the techy stuff, but suffice it to say it is possible for a viewer to capture phonemes (even if they're not transmitted over the network). If the viewer can understand phonemes, then so can the gesture system.

The idea is to add an additional text input field for specifying phonemes. This would allow people to either load custom visemes (in the current RC, blaaaah is viewed as several visemes instead of one constant open mouth), enable "true" voice-activated objects (currently, "voice activated" is used to refer to actions on objects triggered via text chat and the listen event), and some other wacky stuff.

1. Custom Visemes
   Probably more likely with non-english languages, the SL viewer may be unlikely to accurately represent a word with the lip-sync code. Custom BVH animations can be triggered via gestures with either text or hot-keys, but it'd be rather handy for some people to be able to override the lipsync with an animation that accurately represents a particular word.

1. "True" voice-activation
   Add phonemes for "Light on" and "Light off" to a pair of gestures, and have them say the words "Light on" and "Light off" on a specific chat channel, and a user will be able to control an in-world light bulb in SL. Of course, this also has the handy application of being able to echo these back out onto the larger internets via HTTP/XML-RPC and tie them into real-world systems as well :-P

1. whacky stuff:
   1) Adding phoneme triggers to current speech-type gestures so things like "Hi", "LOL" etc can be triggered for the benefit of Residents who don't have voice enabled, but do have speakers/headphones on
   2) Having a custom primitar that accepts text chat input on a specific channel to transform itself into a multitude of objects, adding the necesary phoneme triggers and hey presto! every time you casually say "apple" on voice chat, your avatar turns into an apple! If applied extensively, this does have general emote possibilities, as well as some creative use for standup comedy (i.e. if you say "a horse walked into a bar", your avatar would turn into a horse, turn back to normal and walk, then turn into a bar)