I'm not going to do a thorough review on this, with the 12 version, as I didn't do anything but a rough test on the Asian option.
Again, I "suckered" a store to install this.
Installed quickly and easily.
Didn't crash on the couple of pages I gave it (but didn't send it enough to test it, as it used to guarantee crash with 3 pages or more.
The interface is a little better than it used to be.
Accuracy is horrible (still), but a little better than v11.
The interface is clumsy and difficult to use if you need to alter anything... If you do everything automatically, it's good.
There is no learning for CJK languages... and out of about 800 characters on the first page, there were more than 150 significant errors. It mis-recognized OBVIOUS and simple (less than 8 stroke) characters, would blow punctuation constantly (there are about 30 different "parenthesis" types, it seemed to only recognize 5, and blew the rest, changing them to ").") Full-width and half-width spaces vanished. On a list where there were duplicate lines with just a number change at the end of the line, it would not recognize each line differently, even though it's a clean and well-defined scan with 0° rotation, and monochrome.
One thing it needs to be aware of, that it definitely doesn't, is that most Kanji is fixed-pitched... a space, exclamation point, em-dash, all characters, take up the same width... if it scans this way, you need to choose those full-width characters - not splitting multi-shape glyphs or using proportional Roman text, and if half-width characters are in the mix, it needs to generate those. It didn't do any of this.
While working for a printing company, I had access to a Win95 program called KanjiOCR, it barely ever missed a single character, got spacing correct, and was TINY. It only missed things that were smudged or mis-shaped (scratches, damaged printing plate, etc.) The only issue with that software is that it didn't use Unicode. Why, over 10 years, when tech has advanced (scanners, systems, and pattern recognition), a program that's 12 years old works better than one that's a year old, blows my mind.
It seems that, at least the Roman section of this works much better than it used to. Don't even bother wasting the enormous amount of cash on the Asian option, as you will still have to be HAND-ENTERING a very large amount of Kanji, just to correct it's errors and formatting.
Better yet, try ABBYY, and the unix OCR based free tools still seem to hold up well against the basic version of this software.
I'll try this again with version 13, on the assumption I still haven't found a good Mac Kanji OCR. But, since the Mac software section is vanishing from the only two computer stores in the area, and the Kanji versions don't have a trial version, that probably won't happen.