Book Batcher

PDF → EPUB · experimental

Reflowing a PDF

A PDF does not contain paragraphs. It contains instructions to place runs of glyphs at coordinates on a page. Turning that back into a book means inferring, from geometry alone, where a paragraph starts and stops — and that inference is good on a single-column novel and poor on almost everything else.

PDF reflow needs a careful check

This tool reconstructs paragraphs from the position of text on each page. It works best on simple, single-column documents, but complex layouts can produce text in the wrong order or split paragraphs in awkward places.

Every conversion reports what it measured, including text density, paragraph cohesion and possible columns. Check those results and read the generated EPUB here before relying on it.

Intake

Drop a PDF here

Text-based PDFs only. A scan has no text to extract, and the conversion will say so.

Read this first

A PDF stores placed glyphs, not paragraphs. Where the paragraphs start and stop has to be inferred from the geometry of the page, which works on a single-column book and is usually wrong on anything with two columns, footnotes or sidebars. Every conversion here reports what it measured, and you can read the result before downloading it.

Output format

EPUB — The open standard. Reflows to any screen, and read by everything except Kindle hardware.

Read the result in the reader before sending it to a device. That is not a formality — it is the only way to tell whether the inference worked on your particular document.

What it measures, and reports

  • Characters per page. Near zero means a scan: images of text with no text behind them. The conversion stops and says so rather than producing empty pages.
  • Paragraph cohesion. The share of lines that joined into a paragraph instead of standing alone. Low cohesion means the geometry did not read as prose.
  • Column detection. Two substantial vertical bands of text mean two columns, and reading order across columns is frequently wrong.

When to use something else

If the document is a scan, or the pages are essentially pictures, converting the pages rather than the text is the honest answer — but this build does not rasterise PDF pages, so the tool for that is whatever produced the PDF.

If the document is a two-column paper, a form, or anything with heavy sidebars, read it as a PDF. A badly reflowed academic paper is harder to read than the original, not easier.

If it is a novel or a report set in one column, this usually works well — and the measurements will tell you before you download it.

Protected documents

A password-protected PDF is detected and refused. This does not remove protection from documents and will not try to.