Skip to content
Esc

Japanese image to text (OCR)

Image never leaves this browser

Read Japanese text from screenshots, manga panels, menus and scanned pages. Horizontal text uses the Japanese model; for vertical columns (縦書き), choose Japanese (vertical).

For mixed text such as Hindi with English words
Helps photos of paper; turn off for coloured screenshots

How to use it

  1. Check that the language is Japanese; tick Also read English if the text mixes in English words.
  2. Drop the image or scanned PDF, or paste a screenshot with Ctrl+V. The first run downloads the language data.
  3. Copy the text or save it as a .txt file. Check names and numbers against the original.

Details

Engine
Tesseract 5 LSTM via tesseract.js 7 (Apache-2.0), running in a Web Worker in this tab
Language
Japanese, or any of 120+ Tesseract languages from the list. Mixed-language text: tick Also read English
Input
JPG, PNG, WebP, HEIC, screenshots pasted with Ctrl+V, and scanned PDFs (up to 30 pages, rendered at 200 DPI)
Download
The engine and the language data, a few megabytes each, fetched once from jsDelivr and cached by the browser
Output
Plain text you can copy or save as .txt, with Tesseract's confidence score per image

Questions

Is my image uploaded?

No. Text recognition runs in this tab with Tesseract compiled to WebAssembly. The only downloads are the engine and the language data (from the jsDelivr CDN, cached after the first time); your image is never part of any request.

Does it read vertical Japanese text?

Yes. Choose Japanese (vertical) for top-to-bottom columns such as novels and many manga. Mixed layouts work best when you crop to one direction at a time.

Why are there spaces between characters?

Tesseract sometimes inserts spaces in CJK output. Paste the text into the Word counter, which counts Japanese characters without spaces, or remove them with find and replace.

Can it read furigana?

Small furigana above kanji is usually ignored or misread. Crop it out if it confuses the result.