Glossary
OCR (optical character recognition)
OCR, short for optical character recognition, is software that reads text from an image and turns it into characters a computer can edit and translate. In a manga translator it is the step that reads the lettering inside speech bubbles, before any translation happens.
What OCR does
OCR turns a picture of letters into real text. It looks at the shapes in an image, decides which character each shape is, and returns a string that a program can search, copy or translate. A photo of a page is only pixels until OCR has read it.
Why manga is hard for OCR
OCR was built for printed pages with clean fonts and plain backgrounds. Manga breaks those assumptions. Japanese lettering is often vertical, small kana called furigana sit beside kanji, fonts are stylized or handwritten, and text sits on screentone or on top of the art. Korean and Chinese add lookalike characters that differ by one thin stroke.
Why it matters for translation
A translation engine only sees the text OCR gives it. If a character is read wrongly, the engine translates the wrong word without noticing. Many results that look like bad translation are really bad reading, which is why a large, clean scan helps more than any setting.
How RuneManga handles it
- Japanese uses a reader built for manga, including vertical lettering.
- Korean and Chinese are supported, and their recognition is still improving.
- English, Spanish, French, Portuguese, Indonesian, Vietnamese and Thai can also be read, so comics in those languages can be translated too.
- Auto-detect picks the language from the text it finds; on a page with one or two words it can guess wrong.
- Text outside speech bubbles and sound effects are not translated yet.
Common problems and fixes
- Small or blurry scan: thin strokes vanish and characters are confused. Use the largest copy of the page you have.
- Heavy JPG compression: blocky artifacts look like strokes. Prefer PNG or a lightly compressed JPG or WebP.
- Two pages stitched into one image: send them as separate pages.
- Pale lettering on a tinted background: there is little contrast to read, so expect more misses.
Related terms
OCR comes after speech bubble detection, which finds where the text is, and before machine translation, which changes its language. Furigana and vertical text are two of the reasons Japanese pages are harder to read than printed books.
Does upscaling a small page help OCR?
Rarely. Upscaling makes the image bigger without adding the detail that is missing. A larger original copy of the page helps much more.
Can OCR read handwritten lettering?
Sometimes. Neat hand lettering is often read correctly, while rough or decorative handwriting is easy to misread.