Not being able to read Kanji, even though you know Kanji?


I am far from knowing all the Kanji, but I experienced this with Katakana and Hiragana as well when I was in Japan and now with Kanji.
A lot of writing on shops, in restaurant on TV, etc… are either stylized or hand written. Am I the only one who feel completely hopeless in front of those kind of writings and struggles (sometimes not managing) to read them. Even though you might realized later that you knew all the Kanji or kana use?

I guess my question is: If it true for others, did that become easier with time or are you still struggling with those?

Thanks in advance.



This is an example of what I mean, without the reference, I would never been able to read most of these:


I downloaded some stylized fonts to my computer and use this script to randomly change reviews to one of those fonts. You have to mess with it a bit to get the fonts you want, but it works great.


I can only read styles not deviating to much from the font on this site. That’s because I haven’t made it a priority (yet). I would guess it won’t get easier unless you practise. I don’t think I would be able to read the declaration of independence either.


I use this script as well, and can confirm that it’s fantastic.


Literally just did a double take and went ‘the hell is that?’

@seanblue that looks helpful, thanks for the link


When its stylized, dont focus on just how it looks and instead try and decipher the radicals used. Using that method, I knew all the ones you listed. 言 I just sort of got away with knowing that was the only option. But for…say 首, if you ignore all of the upward sloping lines between the horizontal ones it becomes pretty obvious. It will get easier over time because you will know, for some, “oh there is only one kanji that looks like that” even if you couldn’t instantly recognize it like normal.


Thanks, look like a neat tool, but imagine I’m completely useless when it comes to coding etc… is there a way to download and install this script easily, when I click on the link it gives me a page of code.



The font randomizer is great. I always install good Japanese fonts on my computer when I find them.


When the page of code comes up, the script should load automatically. There are step by step instructions around.


Writing practice helped me a lot. My own bad writing helps me recognize other roughly written Japanese.


I am far from knowing all the Kanji, but I experienced this with Katakana and Hiragana as well when I was in Japan and now with Kanji.

For the hiragana and katakana writings, I’ve been using Real Kana:

Just check all the fonts on the Extra tab, then go to the Hiragana/Katakana tabs to check all the characters you’ve learned, and click the Study tab to begin.

EDIT: I forgot to mention, this misses a lot of the combo katakana used for foreign words. I’ve been supplementing this shortcoming with the Real Kana iOS app

They have a Real Kanji app as well, but I haven’t tried it:


This is where learning the stroke order for the kanji rather than just recognition comes in handy. Despite having some of the strokes merged together, you can still understand the overall direction and it helps make out which strokes they would have been in a neater handwriting.


This is also why those captcha things work so well. Humans learn to recognize similar patterns with enough exposure … computers are just getting there.


In reference to the photo you posted—this is cursive that directly takes its form from the stroke order of these specific kanji. While these forms might look crazy to the average online learner, if you understand the basic stroke order of these characters, you will probably be able to easily understand these (and others) by following along.

Anyway, I just noticed the poster above me said something similar, so here’s something to help you get started on understanding some stroke order if you haven’t touched it before:



It’s just a matter of familiarity.

Yes; I also struggle the first few times I encounter a kanji in a typeface dissimilar from Wanikani (or from a book I’ve learned a kanji from, etc.). When you’re learning to read kanji, a big part of what you’re doing is training yourself to recognize shapes, and when those shapes change, what do you think happens to your ability to quickly assess and categorize it?

Luckily, all it takes is a few encounters with those different typefaces to get your brain to register the similarities rather than the differences. As with anything else, you just have to train yourself, and there’s no way through that but doing.

Now, when they’re calligraphy-based? There are some shortcuts that can help, like learning how different strokes are rendered in calligraphy, but that’s something that will probably take a bit of dedicated study.


Ironically, most captchas you see these days are actually feeding a deep-learning algorithm somewhere. It’s quite a brilliant idea: get humans to tell you what’s in the picture, then compare the responses to other humans for the sake of confirming you’re human, and feed the aggregate responses from humans into a deep-learning algorithm so you can teach a computer what’s in a picture.

Think about that the next time it asks you two questions in a row, even though you’re sure you answered it right the first time. They can give you one image where the answer is well-established previously, and a second one where they need more data. So, they confirm you are human with one question, and get you to assist with machine training via a second question (not necessarily in that order).


I never considered that, but of course it makes sense, @rfindley.
As I was thinking about those captcha things when I posted that, lots of stuff went through my head, like how the ways of weeding out bots has changed in the last 15 years, and how long will it be until we run out of ways to distinguish between a machine and a human.
I have been thinking this in an amateur hobby kind of way ever since I first read GEB in maybe 1980. Self-driving cars and facial recognition that really works are big hurdles that I guess are already here.
So, if machines are learning mostly learning about human communication through social media, they have lots of opportunity to learn. I wonder how long until the big giant hurdle of a bot that can communicate in more than a twittery way with a human that is out to distinguish between machine and human?


Wow, I’d been using the font randomizer that was with the rest of the WK extensions I found, but taking this extra step to drill with wildly different fonts will go a long way towards genuine reading comprehension. Thanks!