PDF to EPUB: how, and why the result disappoints

This is one of those conversions that sounds trivial and is not, and knowing why saves an afternoon of fighting the output.

Why it is harder than the reverse

A PDF records where every glyph sits on a fixed page. It is a description of a printed sheet. It does not know what a chapter is; it knows there is text at these coordinates in this font at this size.

An EPUB is structured text that reflows — the reader chooses the font and the window, and the book repaginates around them. It needs to know what is a heading, what is a paragraph, what is a caption, and in what order they are read.

So converting PDF to EPUB is not a format change. It is an attempt to reconstruct meaning that was thrown away, by guessing from position and size. Going the other way is easy because you are discarding information rather than inventing it — which is why PDF and EPUB are genuinely two different deliverables rather than two exports of one thing.

What converts acceptably

Straight prose. One column, no ornament: a novel, a report, a simple non-fiction chapter. The converter reads paragraph after paragraph, guesses headings from the larger type, and produces something readable.

What survives reasonably well:

• Body paragraphs in a single column
• Headings, when they are genuinely a different size
• Simple inline italics and bold
• Images, as images, though often at the wrong size

What breaks

Everything that made the PDF look designed.

1. Two columns. The converter reads across the page rather than down the column, and the text interleaves into nonsense. This is the most common catastrophic failure.
2. Tables. They arrive as loose runs of text with the rows and columns gone. Tables that survive have to be real tables, and a PDF's are usually drawn lines.
3. Footnotes. They become body text, dropped wherever they sat on the page — often mid-sentence.
4. Running heads and page numbers. Repeated on every screen, because the converter cannot tell furniture from content.
5. Hyphenation. Words broken at the old line ends stay bro- ken, with the hyphen baked in.
6. Drop caps and small caps. Usually a stray single letter, or an all-caps word.
7. Margin notes and pull quotes. Inserted into the flow at unpredictable points.
8. Scanned pages. If the PDF is an image of text, there is no text at all — you need OCR first, and OCR errors then become the book.

How to actually do it

If you have decided to convert:

1. Confirm the PDF has real text. Try selecting a sentence. If you cannot, it is a scan and you need OCR first.
2. Use [Calibre](/blog/calibre-ebook-converter), not a random web converter. It is free, it runs locally so the manuscript does not leave your machine, and its results are as good as anything paid.
3. Set Structure Detection. Point it at the heading pattern it should split chapters on, rather than accepting the guess.
4. Strip the furniture. Calibre's "Search and replace" stage, run during conversion, is the right place to delete running heads and page numbers by pattern.
5. Fix the hyphens. A regex for a hyphen followed by a line break, replaced with nothing, removes most of them.
6. Open it in Edit Book and repair the headings, the contents and anything that landed in the wrong order.
7. Validate it. EPUBCheck errors are what a store will reject on.
8. Read the whole thing. Not skim — read. Converted books fail quietly in the middle.

Budget hours, not minutes, for anything longer than a pamphlet.

The route that avoids all of this

If the PDF is *yours*, you almost certainly have something better than the PDF, and using it is faster than any conversion.

Work backwards to the most structured version you still have:

• The original manuscript — Word, Google Docs, Scrivener, Markdown. Export an EPUB from this. Even a rough Word file beats a beautiful PDF, because it knows what a heading is.
• The layout file — an InDesign document exports EPUB directly, and badly, but with the structure intact.
• Only the PDF — convert, and accept the afternoon.

The reason this keeps happening is that a PDF gets treated as the master copy of a book when it is an output. The master is whatever you can regenerate every format from.

That is the case for keeping the manuscript as plain text: bukpress takes a Markdown source and exports both the typeset PDF and the reflowable EPUB from the same file, so neither is derived from the other and there is never a conversion to do. A correction is one edit and one rebuild of both. Self-publishing from Markdown is the route, and Markdown to EPUB is this job done forwards instead of backwards.

Frequently asked questions

What is the best free PDF to EPUB converter? Calibre. It is free, open source, runs locally rather than uploading your manuscript, and its output is as good as the paid options. Online converters are convenient and send your book to someone else's server.

Why does my converted EPUB look like a mess? Because the PDF had a layout. Columns, tables, footnotes and running heads have no equivalent in a reflowing file, so the converter guesses and the guesses are visible. Plain prose converts far better than anything designed.

Can I sell a converted EPUB? Technically yes, once it validates — and read it end to end first. Converted files are where stores find the reversed tables and the duplicated page numbers, and a refund costs more than the afternoon.

Is it better to convert to DOCX first? Sometimes, and it is not a shortcut. PDF to DOCX has the same structural problem, but Word gives you a place to reimpose real headings by hand before exporting EPUB from that — which is usually the cleanest path for a difficult PDF.

What about a scanned book? You need OCR before anything else, and then the OCR errors are yours to find. For a long scan, retyping the front matter and proofreading the rest is often faster than correcting the conversion.