Internationalizing timed video annotations

Hello,

I thought I'd post here in hopes of getting some direction for an issue I'm having with internationalizing a third-party annotations feature I've developed on top of YouTube's standard player.  The feature allows anyone that embeds a YouTube video to add timed annotations regardless of who created the video.  Take a look here: http://www.embedplus.com/ for a demonstration. You'll see that the tool provides a wrapper that adds a few advanced playback features that the standard embedded player does not currently offer, like the ability jump around chapters, slow motion, instant replay and other things on an embedded video.  
 
Non-US users are beginning to discover the tool, likely through a video-based pronunciation API for English that we're building to demonstrate the enhanced playback options:  http://www.embedplus.com/dictionary/how-to-pronounce-words-and-use-them.aspx .  Perhaps some people here might be interested in this aside.  The idea is that a user searches for how to pronounce a word and the site not only offers audio but also videos of people actually using it in context.  You can checkout these common examples: "How do you pronounce and use gyro (in video)" - http://www.embedplus.com/dictionary/how-to-pronounce-gyro-and-use-it ; "How do you pronounce and use salmon (in video)" - http://www.embedplus.com/dictionary/how-to-pronounce-salmon-and-use-it
 
Just as YouTube videos have built-in automatic caption translations, we'd like to use the Google Language API to provide the same for the typically English text for our annotation feature.  However, an issue is raised that is analogous to the issue of translated HTML button text and other similar confined elements.  That is, the resulting text can consume much more space than allotted.  For important reasons (e.g. to avoid issues from overlaying copyrighted content), our annotations 'box' has a fixed size and location.  This seems to eliminate the applicability of well documented options like text wrapping and elastic positioning/sizing.  
 
Clearly there's the option of distributing extra characters over multiple annotation 'boxes' if translations exceed the fixed space.  The problem, however, is that many annotations are timed to show up at a specific frame and create some effect; hence distributing the text in boxes stretches the text over time which could prevent users from seeing the next effect.  
 
Temporarily pausing a video to "finish" the associated translation for a given frame is pretty much the only option I can think of, but it has a bit of an annoying factor to it.  I then thought to share this issue as it might generate some ideas in this group.  Perhaps it is too obscure or maybe there's a more relevant analog to the issue that I could use to find alternate solutions.
 
Any feedback is welcome.
 
-Tay