Don't worry about the text. In this sort of thing text is functionally irrelevant.It's a combination of alot of things. The images are all resized per the guidelines Realm Works throws at me when it errors. There is alot of images though. Monsters, npcs, locations, overlay maps. Then there is a huge amount of snippets. Of course the database of audio clips probably doesn't help either.
A fairly standard and reliable rule of thumb is a page of plain text in HTML is probably about 1k of storage.
What eats up storage is images and audio. I haven't looked too closely at Syrinscape but if it stores every clip in the RW DB, which it probably has to, that could blow up a realm really fast.