Sorry to bump this - though I reckon it's not too old just yet.
I experimented with this on my local install, and it looks like it's a tad more complicated than it looks because of how the SoundManager library behaves. Energizerspark is correct - currently, the blips are nothing more than a 3-second sound file of "fake blips" that's looped over and over and then stopped when all the text has been typed out (or if there's a pause longer than 25cs). I tried replacing that file with a single blip and make it play per character typed instead of starting it once when text starts typing and stopping it accordingly, but that just ended up playing only a handful of blips per message box since the sound manager was instructed to not let the associated sound overlap itself.
I then changed that flag and promptly had my eardrums destroyed - which made sense, since I'd forgotten to change the amount of loops in the sound initialization, so the blip played once for every letter and also continued to loop for every single instance it was played - so in a text box of 24 letters, just before the end of the box, you'd have 23 blips playing and looping at the same time until they all simultaneously stopped with the typing of the 24th letter. So I set the number of loops to just one per blip (meaning every blip just plays once), but that resulted in a strange side effect that I haven't found an answer to -- the blips would only play once text speed was below a certain amount. They would sync up to the typed letters fine, but once text got too fast (too fast meaning faster than around 0.4x speed - so way slower than the current default speed), they'd just stop playing altogether. And once you approached the threshold, you'd get a couple of blips per text box, but never on the same letters regardled, and very randomly. Also, strangely, changing the amount of loops per blip to 2 would sync them up perfectly with the typed letters regardless of speed, but obviously play every blip twice, which just sounded weird.
I'm thinking this may be a bug with the SoundManager library that AAO uses to manage and play its sounds. As the single blips played and synced up fine when the loop counter was set to anything higher than 1, but otherwise didn't play at all or played very erratically unless the text was set to a slow enough speed, I'm not sure if it could be a problem related to the length of the sound, since it can obviously handle short sounds (or they wouldn't play no matter what the amount of loops was set to). I haven't done any investigating on this (apart from consulting the SoundManager documentation, which said nothing about issues with short sounds), so I'm not sure if or how this can be fixed. But it's definitely a feature I'd be highly interested in having added to AAO, as it would make the typing of text sound much more natural (and the typing of very slow text not sound extremely silly).