Text to speech synched to video

[revolution at duncansoftware.on-rev.com]

I need to have a text file read using the system's text to speech engine
and have it synched to a video. No talking heads just description of the
activity in the video but there would be periods where no speech is
required, in other words, intermitent speech timed to the video. Along the
lines of the old text tracks in QuickTime but what I want is speech
tracks. Does such a thing exist? Anyone have any ideas on how I might go
about  accomplishing this? To complicate things it needs to run on
Windows, and OSX. Linux would be nice but not critical.