# Japanese OCR for Scanned Books: Vertical Text, Furigana, and Reading Order

Convert scanned Japanese PDFs to Markdown or EPUB, choose extraction settings, and handle vertical columns, furigana, and page breaks before translating.

Japanese OCR can turn scanned books into editable Markdown or EPUB files. For vertically printed novels, older publications, and textbooks, the main challenges are column reading order and the relationship between small pronunciation guides (furigana) and the base text. Even accurately recognized characters will produce an unreadable paragraph if columns are stitched together in the wrong order.

If you plan to translate the book, establish its reading order first. A translation can read smoothly even when the extracted text has jumped mid-sentence into an adjacent column.

## Look at a few different page layouts

Open the original PDF and inspect a few representative pages: one with continuous prose, one with pronunciation guides or footnotes, and one with horizontal headings or illustrations. Vertical Japanese prose runs from top to bottom, with columns read from right to left. Horizontal headings, captions, and tables, however, often follow their own layout direction.

Small kana printed alongside kanji are typically furigana—pronunciation guides attached to specific words. Check which word each guide belongs to. If OCR software treats these small kana as standard inline text, the pronunciation may duplicate the word in your output. If the system reads an entire column of furigana before returning to the main kanji, the sentence structure will break apart.

For two-page spreads, inspect the page order and the gutter area. Shadows or tight cropping along the binding can hide characters. Straighten skewed pages and clean up blurred text in the scan where possible. The [guide to preparing poor-quality scans](https://pdfcraft.ai/blog/prepare-poor-quality-scans/) covers these preparation steps.

## Choose Markdown or EPUB

Upload your PDF in the [PDF conversion workspace](https://pdfcraft.ai/pdf-craft/), then select an output format under “Create a result.”

| What you want to do next                                        | Output     | How to use it                                                                                      |
| --------------------------------------------------------------- | ---------- | -------------------------------------------------------------------------------------------------- |
| Correct OCR errors, collect excerpts, or prepare research notes | Markdown   | Unzip the download, edit the text, and keep extracted images together                              |
| Read continuously on a phone or e-reader                        | EPUB ebook | Open the file in an e-reader, adjust font sizes, and navigate chapters using the table of contents |

The reflowed output provides a working version for reading or editing. Keep the original PDF on hand if exact print coordinates, original typefaces, or the vertical layout are critical to your work. For details on managing Markdown downloads and extracted images, see the [complete PDF-to-Markdown guide](https://pdfcraft.ai/blog/pdf-to-markdown-guide/).

## Use refined extraction when you need footnotes

Under “Extraction method,” the standard “Extraction” option captures document content and structure without extracting footnotes. “Refined extraction” refines the overall document structure and extracts footnotes. Choose this option for books where notes at the bottom of the page are critical to the text. This mode uses more credits and takes longer to process; review the [current rates](https://pdfcraft.ai/recharge/) before submitting.

Furigana and footnotes are different structures. Refined extraction does not guarantee an exact visual reproduction of the original pronunciation-guide layout. Compare any small kana against the source scan, especially for personal names, place names, or unique readings chosen by the author. Successfully recognizing the base kanji is not a reason to discard that reading information.

## Keep the Japanese text or add a translation

To proofread the recognized text or take notes in Japanese, turn off “Translate the document.” With Markdown output, you can edit passages while checking against the source PDF. You can also export to EPUB without translation if you want to read the book in Japanese on an e-reader.

To translate into another language, enable “Translate the document,” then select your target language and output mode. “Translation only” works best for uninterrupted reading. “Bilingual” places each translated block directly after the source text, making comparison with the nearby original easier.

Use “Terminology & Translation Notes (Optional)” to define the subject, style preferences, and preferred terminology. For example: “Keep personal names in Japanese and use established English translations for Buddhist terms.” These instructions guide translation choices; they cannot recover characters that are illegible or missing from the scan.

Review your settings and click “Confirm and start.” If you run OCR without translation first and decide to translate later, create a separate conversion task and account for the cost of both runs.

## Follow one complete passage to locate an ordering error

After downloading the result, read through a full paragraph while following the columns on the original scan. If the sentence subject changes abruptly, check whether the text jumped to an adjacent heading, a sidebar, or another page before rewriting it.

For words with furigana, inspect the base text and the small kana separately to see which part was omitted or misplaced. At page breaks, read the final column of the previous page alongside the first column of the next. This helps determine whether an issue stems from a misread character, an omitted phrase, or an incorrect column order.

When logging errors, record the original PDF page number along with a short snippet of the source text. Because EPUB page numbers change with screen and font size, an output page number alone makes it difficult to relocate the passage. For translation-specific issues, refer to the [scanned PDF translation workflow](https://pdfcraft.ai/blog/translate-scanned-pdf/). For Japanese-to-English projects, consult the [Japanese PDF translation guide](https://pdfcraft.ai/blog/japanese-pdf-to-english/).