A weebs guide to translating japanese TTRPGs

Sep 20th, 2020 (edited)
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
  1. Editors Note: Pastebin demanded that I censor myself because this text was considered "offensive". So. Just so you know.
  3. So, you want to translate a game.
  4. If you are reading this, you probably don't know a whole lot of japanese or else this whole thing wouldn't be much of an issue.
  5. Threat not, because actually my japanese is also poor as sh*t and because this is the future or some sh*t, we got tools that will help you to still be able to translate sh*t.
  6. Despite that, it would still help if you know some japanese. You probably do, because let's face it, the fact that you are here probably means that you are at least a little bit of a weeb or otaku or whatever you wanna call yourself, so chances are high that you have at least tried to learn japanese. Anyways, here is a good guide that will help you learn some japanese. https://djtguide.neocities.org/guide.html
  7. It's powered by deep learning and sh*t, and does a pretty good job generally. I say generally, because it sometimes still chokes up and gives you weird translations, but it is far better at it's job than google translate. It sometimes goes a little nuts with names. Just keep that in mind. Like I said, it's getting better.
  8. Next thing you might want to check out is EasyScreenOCR. Links here: https://online.easyscreenocr.com/Home/JapaneseOCR
  9. It is an Optical Character Recognition tool that is pretty decent at understanding japanese, I think because it was purpose build for this type of sh*t.
  10. Alternatively, yandex.com has pretty alright OCR too with translation baked into it. The translation it provides is google translate tier, but as an OCR solution it does it's job well enough.
  11. Also, you might need a dictionary. Jisho is pretty good for that, you can find it here: https://jisho.org/
  12. You might need to upscale images, it can help sometimes. Try using waifu2x, found here: http://waifu2x.udp.jp/
  13. Sometimes, OCR might get confused if the text is literally anything than just a white page with black text on it. You might want to get some image editing software like photoshop, but GIMP or Krita do a good job for free as well.
  14. You obviously also need a word processor. Just take whatever you got. Google Docs works fine in my opinion and allows you to collaborate with people.
  15. It also helps if your OS can display japanese characters correctly. Can't work on sh*t otherwise.
  17. Now, let's get down to business.
  18. First, you need raws. I will not go over how to acquire some here, just go grab some from the trove or other sources, or get your own. Ideally you want raws where you can just copy paste the text, but more often than not, you will probably not be able to.
  19. Thats where OCR comes into play.
  20. You will need images for this, so either extract them from the PDF if you have to (there should be guides and tools for that on the internet about that, so I will not f*ck with that. Just go google that). Foxit Reader, my PDF Reader of choice allows you more often than not to just select a graphic and copy it. Very useful.
  21. Run your images through the OCR. It will give you an archive with some TXT files. They should be numbered.
  22. Alright. Next, you will need to assemble the text. Just match the text in the TXT with the text on the page and copy paste it into a new file of your word processor of choice. Google Docs works well for working cooperatively.
  23. The OCR is not perfect. It will f*ck up sometimes and fail to recognize some characters, especially if they are in some weird font, the quality is sh*t or there are pictures around the characters. You can use photoshop or something to clean the page first so images aren't as much of an issue. Just paint white (or whatever color you need, use your brain for this) over all the pictures so there is only the text left. Black text on white background works best, I'm sure. If the image is poor quality, OCR might not be able to read it. You can try upscaling it. If that doesn't work, I guess you are sh*t out of luck. Sorry.
  24. If you notice that some characters are missing or read wrong by the OCR, japanese skills come in handy. You can try looking up the kanji on jisho with the radicals. Of course, you will need basic japanese skills for that, so better learn up. I learned how to do this kinda sh*t in under a month, so you have no excuse.
  25. Of course, you can forego OCR entirely if you are going for a manual method. By looking up kanji and/or being able to write what you know yourself, you can transcribe the text you want to translate yourself. This, obviously, is incredibly tedious, depending on how well you know your kanji. It might take you minutes at a time to figure out what radicals a kanji might or might not consist of, which means even a few lines might take hours to transcribe. OCR might be kind of tedious as well, but it will in all likelihood speed up the process if you aren't already very proficient with kanji.
  26. I recommend first assembling the text, in a working document, even if you have a PDF where you just can copy the text straight out of the document. It takes some time, but it will probably help you out in the end.
  27. Best to copy paste the text page by page so, say, Page 10 in your working document corresponds to page 10 of the actual book
  28. This is so will have an easier time later.
  30. Once you have assembled the text, you can get translating.
  31. I personally go page by page, sometimes paragraph by paragraph. Simply copy the text and paste it into DeepL.
  32. If there is a break and the paragraph you are translating is on the other page, just copy the rest of the sentence and add it to the text.
  33. Read the translation DeepL spits out. Carefully read it. If something doesn't seem to add up, click on the part that seems f*ed. DeepL will highlight the sentence in question. Hover over the sentence kanji by kanji and see if you can make sense out of it with Raikachamp.
  34. If you are having trouble, it might help to put the sentence into google translate. The translation will probably still be f*ed, but it MIGHT give you a hint on what the sentence means, generally. With that input you might be able to at least roughly translate it yourself.
  35. You can just edit the translation on DeepL.
  36. If sh*t still makes no sense, here are some causes:
  37. - OCR f*ed up and you didn't notice: Go cross reference the appropriate page. Scan for any kanji that isn't what it is supposed to be. If you find some mistake, try to fix it. It is important that the text matches up for obvious reason.
  38. - You might have f*ed up with copy pasting. It happens. You should be able to fix this real quick.
  39. - There might be spaces between kanji/kana. Japanese basically has no concept of spaces, so DeepL f* up often when that happens. Remove the spaces, that often does the trick.
  40. - If there is a break in the sentence, it might f*ck up the translation. Like, if you copy pasted from a PDF you might get linebreaks in a sentence. Simply make sure that a sentence is one continuous line of text.
  41. - It might just be DeepL being stupid. Cut the text out of DeepL and put it back in. Maybe refresh the page and put it back in. If that doesn't work, try to scan the sentence again with raikachamp. It really helps.
  43. Once you translated the passage you wanted to translate, put it in an other document. This will be your translation document.
  44. Same rules as before, try to make it so the page numbers add up, so you don't have to struggle later.
  45. The translation document is just that you have as much of a non-destructive workflow as possible. You might need to go back and fix some translation stuff, so you will be thankful to have the raw text still intact.
  46. I like to just make a copy of the working document and copy/paste and replace in that copy.
  48. That's pretty much it.
  49. Tables can be kind of a pain in the f*ing a*s, but you can use this workflow in tandem with something like MS Excel, or, again, Google Docs.
  50. This whole thing is still pretty new, so this guide might not be completely complete. I'm working on my first translation project right now and I'm quickly making new experiences. I will update this guide whenever I gain some new insight.
  51. You can also hit up the thread and give me some tips, so I might add them here.
  53. Good luck out there anon, let's make it like old /tg/ and get sh*t done.
RAW Paste Data