If you look in the forums, there are models with page-turners: some vacuum-style, some using friction methods. Page-turning is the hardest part, it seems.
It's not certain that the google model works perfectly, BTW:
said by »www.diybookscanner.org/f ··· &p=15144 :
In the video, the presenter said that in testing, the [Google] prototype could scan 60% of books. Of these books, 45% had skipped pages and 45% had folded or torn pages. Hopefully these stats will improve with further development.