I have been pulling my hair out trying to create an immersion workflow to easily immerse in language from videogames. I understand this is not the best study method but I already have an anki daily study and textbooks I do. My issue is that I have problems fitting in time for immersion and this was my solution. I wanted to use these camera and image translator apps to aide with that struggle. Especially since I do not have a large Kanji knowledge base yet and I know enough vocabulary/grammar to work through most sentences using dictionary aides. At a certain point though , that just becomes more study and not immersion. The whole point is to easily understand language and move on to playing the game. I had heard so many good things about yomiwa and deepl, I chose to use them. My experience though has been horrendous.
Image 1 – yomiwa
Image 2 – deepl w/ translation
Image 3 – deepl w/o translation
From my testing both of these apps work fairly well with black text on a white background, written in a traditional format. In Image 1, this sentence is not something I need the app for but it shows just how poor yomiwa performs with text on a background. I don't think I even need to break anything down as it is apparent.
For images 2 and 3, deepl does a much better job actually picking text out of images. Deepl does other things that are even more annoying. The first thing is not shown here in the pictures. It has a habit of taking segmented thoughts, words or phrases and acting as if it is a sentence. For example takes options from a menu and pretends there is a sentence reading "たたかう まほう ぼうぎょ アテム。" which is nonsense. Then tries to infere what this sentence could mean and will give a translation based on whatever it's inference is. And if the game has furigana….deepl goes into a dj session and just freestyles, giving very unreliable translations. Sure, I can just pull the kana from the app and look it up. That just turns this into a study session. Secondly deepl, will regularly miss lines in option menus like ぼうぎょ in this image and incorrectly translates Katakana as you can see it translating ゴブリン to cobbler. The Katakana is mostly a non-issue but just adds more confusion if I miss that part.
Is there any fixes or better apps that other people have used for this sort of thing? I have tried everything from camera images to screen shots. The only reliable way I got it to work is to pull the images into photoshop to put a black box behind the text, which works great if I planned to use them for study or anki cards. It is more work than just working through it with a dictionary though.
I just want to get a quick and mostly accurate translation from an image that has subtitle on a background image. Doing this without fighting through all these quirks is super important as I just want to be immersed in the language but still be able to enjoy playing a game. Any suggestions or input is greatly appreciated!
by Better-Tonight-1457