is there any major difference in the accuracy of these, when it comes to OCR-ing texts?

I'm thinking about getting an OCR program so I can scan some novels and easily make some wordlists for Anki, instead of entering every word manually and looking every Word up on my DS.

I've tried capture 2 text, and while it does a good job and I'm really happy with it, if there is a program that is more accurate, then that's the one I want:p

I've also heard good things about　読んde!!ココ...

I was also wondering, if anyone has used Readiris, whether they've made use of the handwriting recognition feature and how it copes with Japanese handwritten text.... Is it accurate at all?

edit:
apparently I'd be able to get 読んde!!ココ from Amazon Japan... Though I don't even want to think about the amount of tax I'll have to pay:p
and I also see a program called e.typist, which seems a lot cheaper... anyone have any experience with those and would they run on a non-Japan version of Windows?

Capture2text is really good and free, sometimes even better. But it can't multitask like the two other programs, e.g. OCR different areas at the same time or maintain format. I've had no success with Readiris. It's very difficult to setup Omnipage, it took me a while to figure out that it has its own method for scanning Asian text (horizontal and vertical) and it's quite hidden. Anyway, if you are serious about this OCR thing, get Omnipage. If it's a casual thing, then keep using capture2text.

I'd be curious to hear more about OmniPage. Can it handle batch jobs to scan several manga pages at once?

For some reason, both ReadIris and OmniPage have no trial versions, so it's very difficult to figure out if they'd work for me before buying them. You'd think that OCR needs vary so much from person to person (or company to company) that this would be a higher priority.

Netbrian Wrote:For some reason, both ReadIris and OmniPage have no trial versions, so it's very difficult to figure out if they'd work for me before buying them.

There's also ABBYY FineReader — I don't know if it does everything you need but it has a trial version (15-day/50 pages) and there are many reviews that say, in some ways, it's better than OmniPage — less expensive, more accurate, stuff like that (but they usually only try them with texts in English.)

I tried ABBYY FineReader, but couldn't get it to handle manga pages (with vertical text) at all. If anyone has gotten that to work, I'd be really interested.

Netbrian Wrote:I'd be curious to hear more about OmniPage. Can it handle batch jobs to scan several manga pages at once?

For some reason, both ReadIris and OmniPage have no trial versions, so it's very difficult to figure out if they'd work for me before buying them. You'd think that OCR needs vary so much from person to person (or company to company) that this would be a higher priority.

With omnipage, you can batch select all text bubbles in a page. However, if the manga has kanji+furigana you have to be careful not to select furigana areas. If the manga is furigana only though, then you'll get near perfect results. All the error checking doesn't make it worth it. I think your best bet is to find transcripts.

Try e.Typist v.13.0 from the company's website.

http://mediadrive.jp/products/et/index.html

It does an excellent job of JP OCR. You can download a 30-day trial version first. No need to buy a copy from Amazon.co.jp, just buy it digitally. Runs about 15,000 yen for the final version, and works a whole lot better than ReadIris once you get used to using it. I scanned a bunch of stuff with it in Spring, and it generally did a good job, even with mixed JP/EN text. (Although mixed text will cause most OCR programs to barf a bit...)

It's a JP program by a JP company, made for scanning JP text in all of its various formats. I think I posted something about how to use it on the forums here a while back in the OCR thread.

EDIT: Here it is:

http://forum.koohii.com/showthread.php?tid=6542

I tried 読んでココ！, but I never really liked it. Maybe I was too used to e.Typist by then?

SheekuAltair Wrote:With omnipage, you can batch select all text bubbles in a page. However, if the manga has kanji+furigana you have to be careful not to select furigana areas. If the manga is furigana only though, then you'll get near perfect results. All the error checking doesn't make it worth it. I think your best bet is to find transcripts.

Fair enough -- if I'm going to be paying $100+ for an OCR program, I'm not horribly inclined to want to babysit it.

Are there any good centralized repositories for manga transcripts? Either so I don't have to make my own, or if I do make my own, somewhere to put them for others to use?