This is one of the few programs I didn't test drive from macupdate or versiontracker. I actually convinced a mac salesperson to let me try this on one of the G5s in-store. After all, they had a full setup with scanner, digital photography, and a HDTV -- everything. Installation was initially scary, as it looked more like i had accidentally installed a bootleg with virus. But after being assured they didn't care (the unit was protected), we installed and input the password. Seemed simple enough. Had a few basic issues, like reinstalling the TWAIN driver from the scanner's site.
The the problems came. Whoever designed the menu system and the language (words) used, seriously need to go back to school and to learn about precise terminology. It was very confusing, and took us both quite a lot of time to figure out what was meant by certain terms. And on top of that the "search" function in the "help" file doesn't search, it provides an "index." Which is pretty useless if you're tyring to find an explanation for some terms, or trying to find out why something isn't working the way it should be under a Mac interface.
After all that, I scanned 5 pages from a paperback I had on me... very simple document. It managed to outline the text, but in weird shapes, when a simple rectangle would definitely do (there is no option to force this, so it goes about making weird "polygons"). Attempting to "reshape" the polygons had the whole giant shape misaligning everywhere, it crawled about like some form of amoeba. Definitely bad. We eventually let it just do it's thing, and clicked analyze. (You can "redraw" the analysis boundaries, but it requires a lot of mouse-hopping to get to make a simple rectangle, which is contrary to the simplicity the automation is supposed to provide).
A very simple font was used (Goudy Old Style), with a few exceptions (like diagonal hyphens), the text should have been simplicity itself to recognize. Nope, the self-learning process couldn't seem to recognize double and single quotes properly, and even after being forced to learn several of them, it started turning them into uppercase "H"s. Not only that, it kept changing double lowercase "o"s into double zeroes. It couldn't recognize periods, colons and semicolons, and when there was noise between lines, instead of ignoring that, it dwelled on it making weird guesses. The worst was when it started splitting double quotes into singles before we finished the five pages (and it did this with a couple of other characters, like a "u". There's no way to tell the program it's totally screwed up, so you have to either "delete" the supposed character, or stick something in under "don't learn". Either way, you have to go back later and correct it in the output.... if you can get that far.
In the middle of the fourth page, the program crashed. It lost all the settings and the "learning."
We tried it twice more, once with TWAIN, and then with pre-scanned pages. Crashed each time, before we finished, but in different places.
Finally tried just a single page. It managed to get through that, but with a large number of errors, which meant we had to go back and "proofread" it. I could have hand typed the page three times by the time we got through the proofing.
On top of that, we found that there is no way to keep paragraph formatting (as opposed to single lines with a hard-CR at the end), and maintain things like line ending hyphens... it removes them whether or not they are used normally or as a line/word split.
I was kind of hoping this might be my answer to a program that could OCR Kanji, but with this performance, there's no way I would attempt it.
Besides, the version in the store was apparently from 2005 even though it was brand new there. An 11.5 version is out but I don't see it here. However, if you go to their site, you think you'll get a test download, but it does nothing but take your e-mail addy and then send you to more advertising. There's no tech specs on the asian upgrade, and attempts to find any, just send you back to the "test" version page, which only has "video" clips.
The salesman (cute guy, BTW, not a total waste of time) let me take the software home to test with a deposit. It performed even worse on my G4.
The company claims it is the best Mac OCR program... Of course it is, it's apparently the only one with a GUI frontend that I can find.
What really blows my mind is that since Omnipage on the Mac died a while back, pattern recognition has gotten much more sophisticated, but I don't see it here, at all. One would think that the program would be able to take a font file and read it's metrics and splines in order to train itself, making it a whiz to use, if you have the typeface. No, it's still working on making rough guesses (and bad ones at that).
This program also requires you to have a TWAIN interface... It doesn't have any real scanner interface of it's own, so if you have an old scanner (read OS9 or earlier) you can't use it, TWAIN doesn't work across the "Classic" barrier.
This program performed horribly, in all taking more than 5 times on average to do the work, than it would require me, or even an average secretary to type the work by hand. Kind of pointless and a waste of time and money. I'm hoping I can find a used copy of OmniPage Pro X... it worked better when I was using it on my old Mac IIfx. Though that won't solve my issue in trying to find a kanji OCR program.
I took the software back, and I hope the sales-guy tells the store to return the stock.
I am so glad I didn't buy this stuff on-line, I'd be out $400, and may not have even gotten actual software, based on my and other's experience with the company. (This is why I love MacUpdate)
Maybe by version 20, this might be working well enough to seriously use. Errrr. Maybe not, just checked the date on this "update", Oct 31 2005... Happy Halloween... guess I found the monster.