Hi, Mike.
Just tried New OCR engine - cannot see a big difference in skipping non-hanzi chars.
The engines are a bit different, but it seems the new one makes even more mistakes in distinguishing that kind of chars (even for clear readable pinyin with tones).
PS: I'm still asking just to make the green box movable
The main purpose of this functionality is not to read "real" Chinese books, but to read a learning content - tutors, special texts with Pinyin, bilingual texts, etc. Usually this kind of content has repeated blocks, i.e.:
__
a string of hanzi
a string if pinyin
hanzi
pinyin
hanzi
pinyin
...
__
So it's easy to adjust the box once and just to move it along a page, skipping non-hanzi strings and putting on Chinese chars only.