EPUB vs PDF

The one difference

A PDF is a page description. It says: draw this glyph, in this font, at these coordinates, on a sheet this big. The line breaks and the page breaks are baked in, and they were chosen before the file left the machine that made it. That is why a PDF looks the same everywhere, and why it is unpleasant on a phone.

An EPUB is a zip of XHTML and CSS — a small website — with the layout deliberately left unfinished. When you open it, the reading app does the typesetting itself, in your font, at your size, on your screen, with your margins. That is why it is comfortable on a phone, and why nobody can tell you what page you are on.

Neither is a better piece of engineering. They answer different questions: what should this look like versus what does this say.

What follows from it

The row worth dwelling on is the accessibility one, because it is the least known. An EPUB is marked-up HTML: a heading is tagged as a heading, a list as a list, the reading order is explicit. A screen reader can navigate it the way it navigates a web page. A PDF may carry the same tags and very often does not — an untagged PDF is a bag of positioned glyphs, and reading it aloud in the right order is guesswork.

Converting is not symmetric

Most comparisons stop before this, and it is the practical half. The two directions are different kinds of problem, and one is much harder than the other.

EPUB → PDF: the pages must be invented

Nothing in the EPUB says where a line ends. Something has to choose — every line, every page — and how well it chooses is the entire quality difference between converters.

Here is what that looks like as a number. Breaking lines greedily to fit, with no hyphenation, leaves about one line in five visibly loose in a justified novel — 17.5% for Pride and Prejudice, with a further 3.3% of lines that could not be fitted at all. Adding hyphenation, using the same algorithm TeX uses and the pattern set for the book’s own language, brings that to 6.0% and 0.6%. Same book, same size, same typeface.

It also moved the page count from 415 to 411. That is the point worth taking away: there is no correct page count to preserve. A conversion does not recover the book’s pages, it manufactures a set, and a different setting manufactures a different set.

Two other things are genuinely lost, and are worth knowing before you convert rather than after:

• Reflow, permanently. The PDF is now a fixed page size. Wanting larger text later means printing on larger paper, or converting again at a different setting. If the reason you want large text is eyesight, pick the large-print style at conversion time — 16 pt, generous leading — rather than trying to zoom afterwards.
• Scripts the PDF standard fonts cannot draw. The fonts built into the PDF format cover the Latin alphabet and little else — no Cyrillic, no Greek, no Chinese, Japanese, Korean, Hebrew, Arabic or Devanagari. Drawing those means embedding a real font in the file, which is why most converters simply drop the characters. Ours embeds one: a book with Greek or Cyrillic in it fetches a proper book face and sets every character. For the scripts even that face does not cover, it still checks before you press anything and says so, rather than returning a file of blank boxes.

PDF → EPUB: the structure must be guessed

Going the other way, you are not adding information — you are trying to recover information that was never recorded. A PDF stores glyph positions. It has no concept of a paragraph, a heading, a footnote or a column. All of that has to be inferred from coordinates and font sizes.

We measured how thin the ice is while building a PDF parser. Numbered lists come through cleanly, because the digits are real text in the file. Bulleted lists vanish — the bullet is drawn by the layout engine and is typically not in the text layer at all, so the items arrive as unmarked paragraphs. Headings have to be detected by measuring font size against the median. Two-column pages read across the gutter unless something notices the columns. Words hyphenated at a line end arrive hyphenated.

This is why PDF-to-EPUB output is usually disappointing, and why we do not offer it. If what you want from a PDF is its text rather than its pages, PDF to Markdown is the honest version of the same job: it recovers the structure it can actually see and does not pretend to have found the rest.

The converters that screenshot a reader

One practical warning, because it decides whether the PDF you get is any use. The easy way to write an EPUB-to-PDF converter is to render the book in a headless ebook reader and photograph each screen. It sidesteps typesetting entirely.

What you get back is a PDF full of pictures. No text in it: search finds nothing, you cannot select a sentence, a screen reader has nothing to read, and the file is many times larger than it needs to be. It looks fine until you try to use it.

Typesetting properly is not slow, which is the thing that makes the shortcut hard to justify. The Pickwick Papers — 1.75 million characters, 974 pages once set — takes about two and a half seconds and comes out at 2.3 MB, with every word selectable.

Two cases where the question does not apply

• Fixed-layout EPUBs. Comics, picture books and many textbooks ship as EPUBs that forbid reflow — one designed page per document. That is a PDF that took the scenic route, and converting it gains nothing.
• DRM'd books. A book bought from Kobo or Google Play is an EPUB whose contents are encrypted against your account. It unzips and it is still unreadable, and no tool running in a browser tab can change that. Kindle books are not EPUB at all. More on both.

Questions

What is the difference between EPUB and PDF?

A PDF describes finished pages: this glyph, at these coordinates, on a sheet of this size. An EPUB describes text and leaves the pages undecided, so the reading app breaks the lines and pages when you open it, using your font and your screen. Everything else people list — reflow, file size, page numbers, accessibility — follows from that.

Is EPUB better than PDF?

For reading prose on a screen, clearly. For anything that must be printed, must look identical to everyone, or depends on alignment — forms, contracts, sheet music, textbooks with figures — PDF wins, and not narrowly. They are not competing at the same job.

Which is smaller, EPUB or PDF?

Usually EPUB, because it stores no fonts and no layout. Converting Pride and Prejudice from a 545 KB EPUB gives a 1.11 MB PDF of 408 typeset pages — measured, August 2026. Both are small; a converter that hands you a 20 MB PDF of a novel has screenshotted an ebook reader rather than typeset the book.

Does converting EPUB to PDF lose anything?

The text survives intact and so do the illustrations. What changes is that pages get invented — something has to choose every line and page break, because the EPUB never contained any. You also lose reflow permanently: the PDF is now a fixed size, and making the text bigger means making the sheet bigger.

Can I convert a PDF to EPUB?

You can, and it is the much harder direction. A PDF records glyph positions, not structure: it does not know what a paragraph is, and bullet points are typically not in its text layer at all — we found this measuring our own PDF parser, where numbered lists came through because digits are text and bulleted ones vanished because the bullet glyph is not. Everything a good EPUB needs has to be guessed back.

Which one should I email someone?

PDF, unless you know they read on an e-reader. It opens on everything without an app, it looks the same on their screen as on yours, and nobody has ever been confused by receiving one. EPUB is the better format and the worse attachment.

Line-quality, page-count, file-size and timing figures are our own, measured across a 14-book corpus while building the EPUB to PDF converter in August 2026. The PDF text-layer findings come from our own pdf.js parsing work, July 2026.