Skip to content
Droid Sentence MinerDroid Sentence Miner

Books and manga

The Book tab is for paper: light novels, manga, textbooks and anything else that is not on a screen. A photo is read with ML Kit, laid out as Japanese page text, and overlaid on the image so the printed characters themselves become tappable.

Take a page photo

Section titled “Take a page photo”
  1. Open the Book tab.
  2. Tap Take photo and photograph the page flat and well lit, filling as much of the frame as possible. On Samsung phones this opens Samsung Camera; elsewhere it opens any camera app that handles image capture.
  3. Wait for the page to be read. Paragraphs are outlined on the photo.

The photo is decoded upright (EXIF is applied) and capped at 3000 px on the long edge. No camera permission is needed because the camera app returns the image.

You can also pick a photo from the gallery with the image button, or share a photo from another app to Droid Sentence Miner: it opens in the Book tab and is read right away.

Layout

Section titled “Layout”

Book mode is not a generic OCR field. It understands Japanese page layout:

  • Tategaki columns are ordered right to left, then top to bottom.
  • Horizontal lines are ordered top to bottom.
  • Kana-only ruby lines and very thin lines are dropped. Ruby is detected by width relative to the median column, after OCR text fixes.
  • Page numbers and non-Japanese blocks are ignored.
  • A block that ends mid-sentence is joined with the adjacent block.
  • …… is restored after Unicode normalization (NFKC turns it into dots).

Mine from a paragraph

Section titled “Mine from a paragraph”
  1. Tap a paragraph on the photo to open its text.
  2. Tap a word to open the definition popup. Tap Japanese inside a definition to look that up too; the back arrow returns.
  3. Tap Anki. The card image is the paragraph cropped from the photo, and the sentence is cut at 。！？ while keeping closing quotes.

Hold and drag over the text to force your own selection. A hold without dragging selects one character, which is useful for a single kanji. The ✎ button corrects OCR text before mining; word positions are preserved for the parts that did not change.

Read on the photo

Section titled “Read on the photo”

After a photo the page opens in an overlay reader:

  • the OCR text sits invisibly over the printed characters;
  • tap any character to look up the word starting there;
  • hold and drag to select characters;
  • pinch to zoom and drag to pan;
  • ✕ or Back returns to the paragraph view.

This is the closest thing to tapping a real manga page. The page stays alive across tab switches, so you can leave the Book tab and come back without taking another photo.

Limitations

Section titled “Limitations”

Book photos always use ML Kit. meikiocr is a horizontal game-text reader and is not used for tategaki pages. ML Kit can still misread stylized manga fonts; use ✎ before mining.

Section titled “Related”