need to translate japanese text from images. specifically: game screenshots, photos of menus at restaurants, and the occasional manga panel.
tried google translate’s camera feature and it works okay in real-time but falls apart on complex kanji. haven’t found something that handles the full range well.
what tools actually work for image-based japanese translation? specifically looking for: something that handles both printed and handwritten text, works on screenshots not just live camera, and ideally handles vertical text layouts since a lot of japanese content uses those.
been relying on manual lookup for anything complex which takes forever. looking for the tool that people who actually work with japanese content use.
DeepL for Japanese translation quality is genuinely noticeably better than Google for anything beyond simple sentences. Game text, narrative writing, and nuanced expressions translate more naturally. For casual signage Google is fine but for anything you want to actually understand deeply, DeepL is worth the extra step.
google lens being integrated into google photos on android is convenient for translating screenshots. you screenshot something, open photos, tap the photo, tap lens, select translate. the whole flow takes about fifteen seconds and it handles japanese vertical text orientation correctly which is the main challenge.
PowerToys Text Extractor for desktop is the PC workflow that most people don’t know about. Screenshot a region with the OCR tool, copy the extracted text, paste into DeepL. It handles Japanese well and works for any application including games in windowed mode. Free, no installation beyond PowerToys itself.
for japanese image translation, the right tool depends on the use case:
real-time camera (restaurant menus, signs): google translate camera feature works well for printed text. point, it scans, it translates live. handles horizontal text reliably, vertical text less so.
screenshots and static images: google lens (available in the google app or photos app) handles screenshots better than the camera feature. upload the image, select text, translate. handles both horizontal and vertical layouts reasonably well.
manga and complex layouts: google lens works for most cases, but for dense manga with overlapping text and speech bubbles, dedicated tools do better. Mokuro (open source) processes manga pages specifically. for casual use, the manga reader app on android has built-in OCR and translation.
handwritten text: google lens has improved significantly on handwritten Japanese but still struggles with cursive (草書/行書). if it fails, try Microsoft’s lens app which uses a different OCR model and sometimes handles handwriting better.
for game screenshots with stylized or pixel fonts: neither tool handles pixel art fonts well. your best option is to manually look up characters using a kanji drawing tool like jisho.org’s draw feature or the Google Japanese input keyboard’s handwriting mode.
for regular printed Japanese text in clear images, google lens is the practical answer for most use cases.
The vertical vs horizontal text handling being a specific challenge for Japanese OCR is an important distinction. Most Western OCR tools assume left-to-right horizontal text and fail on vertical Japanese text. Google Lens and dedicated Japanese OCR tools handle both orientations, which is why they’re better for this use case than general OCR tools.