Here's my suggestion: If there is a speech pack available, before every line AGS calculates the following thing:
The frames to be played(syncronized) / The length of the sound file = How fast should the frames go
This would open up new possibilities - real-time sync speech like in KQ6, for example.
While this is possible, it wouldn't really help since voice speech doesn't have a constant speed with which each letter of the sentence is said - some words are spoken faster than others, there are pauses and so forth.
If you want voice lip sync, get 2.55 beta and read the bottom of the Lip Sync section of the manual - but bear in mind it's an unofficial feature and no support is offered.