1. the book is scanned so we have one image per page
2. each page’s image is cut up so that each word goes into its own image so we have as many images per page as there are words in it.
3 the book in Topaz format is nothing else but container for those words’ images.

Having each word in separate image allows for reflowing (by rearranging words/images) and resizing (by resizing the words’ images)

You can not change the font because there is no concept of font in this format – you see the font that was used in printed book. Hence no font embedding – there are no letters to apply the font to.

4. in addition the book is OCRd. The result is stored as _hidden_ information that is used for instance to search the text of the book – you never see the result of the OCR directly.

This was the first format that was developed for Kindle. This is what was shown to publishers when Amazon was just thinking about the whole e-reader business and had a prototype of the device. The main reason was that the conversion from printed book to a e-book which looked almost like the original was fully automated.

Later MOBI became the principal format for Kindle, but to have the book in this new format it basically has to be redesigned almost from scratch (to have all the hyperlinks, etc in place). Hence MOBI came later when publishers got on board.

All that was not a speculation, I just repeated what the developer said about his work.

Basically, it looks like a photocopy of a print book; that is, it doesn’t look nice. You tend to end up with uneven fonts, and ink splotches on the page. Topaz is one of the reasons I check the sample of every Kindle book prior to purchase. Much of the formatting done for the Kindle has been extremely shoddy.