Apologies if this has been asked like a thousand (ooo I know that kanji, 千) times; I’ve had a quick search but I’m tired atm.
I use the browser plugin Rikaichamp for helping with reading, is there something similar when looking at digital images (such as manga)? I don’t want something that translates everything, but otherwise I’ll have to zoom in and translate with Yomiwa app on my phone (which is a bit fiddly)
Basically I’m wanting to start basic manga and looking for some support; I’m hoping that digital manga will enable me to zoom in and see characters more clearly than print.
Sorry for the stream of consciousness, hope it made sense.
OCR is generally a pain in the ass, but I’ve actually been working on a small project for my own manga reading pleasure. Right now it’s hooked to an online translator but I’m planning to extend the interface eventually and steal borrow rikaikun’s offline database.
It’s not always working and can be iffy about the borders/fonts but overall it’s more or less achieving its goal.
I’m using KanjiTomo, and while it does make some parsing mistakes, it’s very convenient as an OCR dictionary.
I too wanted the ability to zoom in to see the characters clearly, but bear in mind that how far you can zoom in and still have a clear enough image will depend on the specific manga image quality. Several times I happened to zoom in just to be presented with an unrecognizable blur. In such cases, a printed book and a magnifying glass might work better.
If you don’t mind working in two tabs, there’s also Copyfish for OCR, and Yomichan for dictionary.
Click Yomichan’s search icon to open a tab, enable clipboard monitoring, do OCR in Copyfish (not always the best result, but typically fairly good), then click on the copy to clipboard button. Yomichan reads that and auto-shows dictionary entires for all the words.
I’ve tried all methods mentioned above and japanese.io gives me the best experience so far. The only downside is 100$ per year, so I hope I will learn enough kanji to stop using it one day.
I can’t get Yomichan’s shift+hover to work with Copyfish results on a local file (haven’t tried with a non-local image), but I definitely use it for web pages and e-books. (I haven’t tried Rikaichamp, so I don’t know how it compares.)
I tried Copyfish and it was nearly useless for me, it recognized barely one character in a whole bubble, if that. I wonder if I did something wrong. Tried with handwritten text first, feeling lucky, then with clear printed text, still nothing. Setup is Windows and Chrome, don’t know if that plays a role. Kanjitomo on the other hand has been fairly consistent for me, with only very occasional fails.
On Copyfish, where it says “OCR Result”, does it says “(Japanese)” underneath? I found that when I had the worst results, it had changed back to English.
It doesn’t always give the best results, but it should be at least somewhat decent for digital manga pages. (I don’t have experience with other usages.)
I mention the language setting because that’s what led to my having bad results with it early on. Regardless, in the end, we’ll each settle on what work for us =D
Indeed it was the language setting that was causing problems. How silly of me! Thank you.
How is its accuracy for you？Do you need to zoom in a lot for it to be accurate?
I don’t do OCR enough to give a proper answer, but I’d say maybe 70% to 98% accuracy, depending on the text’s font and resolution.
I wanted to bring up some examples where it’s failed, but in trying a handful of manga, I seem to have not be able to get an OCR error. It’s not flawless, but it seems like my results are a little better than in the past. Perhaps they’ve improved things a little? (There are punctuation errors shown here, though.)
a snap judgement from purely looking at the website’s landing page, but it looks like it’s the same as Rikaichamp - a browser plugin that translates highlighted/mouse-over’d text…
