PDF を EPUB に変換

このツールでできること

The reason people want this is specific: a PDF on a phone or an e-reader is close to unreadable. The page is a fixed sheet of A4 and the screen is not, so you either zoom and pan or squint at six-point type. EPUB re-flows to whatever width it is given, and that single property is the whole value of the conversion. It also means the original pagination is discarded - the text is rebuilt as a stream of chapters, paragraphs, headings and lists, not as a copy of the printed page.

Use it for anything long that you intend to read rather than consult: a book or manual that only exists as a PDF, a lengthy report you want to get through on a commute, a thesis, a set of papers going onto a Kobo. It is the wrong tool for anything where the layout is the content - a form, a score, a plan.

仕組み

1. Drop the PDF onto this page. It is checked for a text layer, because there is nothing to re-flow in a scan.
2. Choose where the chapters come from: bookmarks, detected headings, or page boundaries as a last resort.
3. Fill in the title and author if the document's own metadata is wrong or missing - this is what an e-reader shows in its library.
4. Under advanced options set the language if the document is not in the language of this page, and narrow the page range to skip front matter you do not want.
5. Press Convert to EPUB, then send the file to your reader however you normally do.

Chapter detection is the feature, not a detail, because an e-book without chapter breaks is one long scroll with no table of contents and no way back to where you were. Bookmarks come first: a document's own outline is its author's statement of where the chapters are, and it beats any heuristic. Heading detection is the fallback, promoting short lines set in larger type. Page boundaries are the last resort - not really chapters at all, but at least somewhere to jump to.

Setting the language matters more than it looks. An EPUB carries a language tag, and readers use it to decide how to hyphenate and how a screen reader pronounces the text. Left on automatic, the tag follows the language of this page rather than the language of your document, which is a sensible default and the wrong answer if you are converting a Spanish book on an English interface.

Title and author are worth a moment even though they are optional. Left empty, they are taken from the PDF's own metadata, which is frequently the name of a template someone used in 2011 or the path of a Word file. That is what appears on your reader's shelf, next to a book with no cover.

Everything runs in your browser: the text extraction through PDF.js, the EPUB packaging through a ZIP writer in a Web Worker. No part of the book is uploaded, which is the honest answer to the question people ask about converters that handle copyrighted material.