What about the idea using shared objects instead of cuepoints for captions?
The clients will see the captions faster than with cuepoints but on the other hand the captions will be lost unless I save it on the server in a separate file. I use different shared objects for different languages.
Do u think that this is feasible?