Convert EPUB to PDF

An EPUB has no pages to preserve

This is the fact everything else on this page follows from, and almost nobody states it plainly. An ebook in EPUB form is not a document with a layout. It is a zip archive holding chapters written as XHTML, a stylesheet or two, the illustrations, and a small catalog file that lists what is inside and what order to read it in. There is no page size recorded anywhere. There is no page break. The text is meant to reflow — to rearrange itself around whatever screen and type size you happen to be reading on.

For an ordinary archive where the job is to inspect paths or retrieve one entry, use the local ZIP file viewer. This converter instead interprets EPUB-specific reading order, chapters, and resources.

So when a reader shows you page 47, that number was invented a moment earlier by that device, at that type size, in that orientation. Turn the font up two steps and page 47 becomes page 61. Hand the same file to a phone and the whole numbering changes again. There is no original pagination sitting in the file that a converter could faithfully reproduce, which means any tool promising to preserve the book’s layout is promising something the format does not contain.

EPUB and PDF, in brief

EPUB

Full name	Electronic Publication
Extension	.epub
Format type	A reflowable ebook — a zip of XHTML with no fixed pages
MIME type	application/epub+zip

PDF

Full name	Portable Document Format
Extension	.pdf
Format type	Fixed-layout document
MIME type	application/pdf

What this EPUB to PDF converter actually does

It sets the book again. The archive is unpacked, the catalog is read to find the reading order, and each chapter is measured and laid onto paper you specify — Letter or A4, your margin, your type size — with headings, paragraphs, lists, block quotes, tables and the book’s own illustrations carried across. Chapters start on fresh pages. Page numbers read Page 3 of 214 rather than a bare digit, because a converted book is the sort of thing people print and then drop.

What it does not do is apply the publisher’s stylesheet. Doing that properly needs a complete HTML layout engine with pagination, which is a piece of software the size of a browser; doing it halfway — a handful of CSS rules interpreted on a guess — produces a result that looks deliberate and is not. Between an approximation of someone else’s design and a clean setting of the same words, the second is the honest product, and it is also the one you can adjust.

Reading order comes from the spine, not the file names

Inside the archive the chapters are usually named something like part0012.xhtml. Sorted as text, part0012 lands before part002, forewords scatter among the middle chapters, and some publishers name files after their content rather than their position. A converter that walks the archive alphabetically produces a book whose chapters are shuffled — and the result looks perfectly fine until somebody reads it.

The true order exists in exactly one place: the <spine> element of the package file, a list of references pointing into the manifest. That is what gets followed here, and the test suite proves it with a book whose spine order is deliberately not its file-name order. The table of contents is a separate thing again, and it is not used for ordering — a book is allowed to contain chapters its contents page never mentions.

Locked books, and why the lock stays on

Ebooks bought from most stores, and everything borrowed from a library, arrive wrapped in DRM. Adobe’s scheme encrypts the chapters against an Adobe ID and leaves a rights.xml in the archive; Apple’s Books store uses its own and leaves a sinf.xml. This tool looks for those before it does anything else, names which one it found, and stops. It will not decrypt them — not as an option, not behind a warning. That conversion belongs in the reader authorized to the account that bought the book.

There is a subtlety here that catches most converters out, and getting it right is the difference between a cautious tool and a good one. An encryption.xml in the archive does not always mean DRM. Publishers routinely scramble their embedded typefaces so the fonts cannot be extracted and installed, using a method the format specification itself defines — and a book treated that way is completely readable. A converter that merely checks whether the file exists refuses those books, confidently and wrongly. This one reads which algorithm was used and what it covers, so a book with obfuscated fonts converts normally and simply notes that the publisher’s typefaces did not come along.

Kindle files are a different format entirely

.mobi and .azw3 are not EPUBs and cannot be read as one. They are PalmDOC containers in Amazon’s own record format, with their own compression — a completely separate specification that happens to hold a book too. Drop one here and you will be told what it is rather than shown an unhelpful broken-archive error, because the check reads the actual bytes and not the extension on the end of the name. Calibre converts them on your own computer, free, and its output opens here.

Calibre deserves the mention in its own right. It is free, it has done this job well for years, and it does considerably more than this page does. The argument for a browser tab is narrower and worth being honest about: you have one book, you want a PDF now, and installing a full library manager to get it is out of proportion to the task. If you convert books regularly, install Calibre.

Which languages come out right

Every PDF reader can draw fourteen typefaces without being sent any font data, and between them they cover Western European text. Books in English, Spanish, French, German, Italian, Portuguese and the Nordic languages set correctly with nothing extra downloaded. An optional font, offered only when the book needs it, adds Cyrillic and Greek.

Chinese, Japanese and Korean do not, and neither do emoji: those need font files in the tens of megabytes and the tool says so before you wait rather than after. Arabic and Hebrew are refused on purpose even though the letters exist, because setting them requires reordering and reshaping the text, and letters drawn in the wrong order look entirely correct to anyone who cannot read the script. A visible substitution is the safer failure. Whatever gets swapped is counted and named on the receipt, so you know exactly what changed rather than discovering it on page 90.

Where the book gets read

The archive is unpacked, the chapters are parsed and the PDF is assembled by code running in the tab in front of you. What makes that worth insisting on here more than elsewhere on this site is what an ebook is: usually bought, sometimes borrowed from a library under your own name, and always something you chose to read. A reading list is a more revealing thing to hand a stranger than a holiday photo, and the safest way to keep that private is for no copy to be made in the first place.

That claim is checkable rather than promised. Open your browser’s Network panel, drop the book in, and watch the request list stay empty of it; the fuller procedure, including the test most people skip, is written up in how to verify a tool runs locally. Hand the book in first — that one act is what fetches the PDF writer, and it is the last thing this address is asked for — then take the machine off the network and convert the rest of the shelf anyway.

One kind of file has to be dealt with before it gets here, and it is worth naming because the error would otherwise be baffling. A Kindle book is not a zip archive at all — it is a Palm database, a container design from 1996 with numbered byte ranges instead of filenames — so handing one to a page that opens archives produces a complaint about a broken archive, which is true and useless. Both Kindle formats now have their own address: MOBI to EPUB unpacks the older .mobi, whose text sits in one continuous stream with page breaks as its only structure, and the EPUB it produces is the file this page wants.

The newer format takes a different route to the same place. AZW3 to EPUB works on .azw3 files, where the chapters are still separate XHTML documents underneath and come back out with their titles attached — which matters here, because a chapter with a name gets a heading in the finished PDF and a chapter without one gets a number. Neither page removes DRM: a book still tied to a Kindle account is identified as locked and left alone, the same treatment Adobe's lock gets above.

Once the PDF exists it behaves like any other. Several books or chapters can be merged into one file in the order you choose, and if you want the words back out as plain text rather than pages, PDF to text goes the other way. Should you decide the fixed page was the wrong home for the book after all, converting a PDF to EPUB takes it back to a reflowing format — dropping the page breaks, running heads and page numbers that a printed layout needs and an ebook has nowhere to put. Documents that started life as notes rather than books have their own route in Markdown to PDF, which uses the same typesetting engine as this page and offers the same paper controls.