Convert PDF to ePub

Turn a PDF into a reflowable ePub 3 e-book for e-readers. Best on text-based, single-column PDFs. No signup. Runs on our server; files are deleted automatically within about two hours.

Drop your files here

.pdf · up to 200 MB · up to 30 jobs an hour · uploaded to our server and deleted after processing; the result is wiped within about two hours

Choose files

Auto-detectSingle columnTwo columns

More settings

Run OCROCR languageInclude imagesTitle (optional)Author (optional)

How to convert PDF to ePub

1
Upload your PDF
Up to 200 MB, up to 400 pages. Password-protected files are rejected before upload.
2
Pick column layout and language
Auto-detect columns or force single-column reading order; turn OCR on for scans.
3
Download the ePub file
A standard ePub 3 file that opens in any e-reader app.

Why the restructuring is only approximate

A PDF records where each character sits on the page, as x and y coordinates, and not the headings, paragraphs or reading order. Turning that into ePub's flowing text means Calibre has to guess the document's structure from those coordinates: which lines belong to the same paragraph, where one column ends and the next begins, which text is a heading rather than body copy. It gets this right most of the time on ordinary documents, but it is a best-effort reconstruction, not an exact translation, and it is better to know that before you convert than to find out from the result.

Which PDFs convert best

A single-column document with a real text layer, such as a novel, a report or most books, reflows cleanly. Multi-column layouts, tables and running headers or footers are the cases that go wrong. A two-column scientific paper can end up with its columns interleaved rather than read left then right, which is what the “single column” setting is for: it tells Calibre not to try detecting columns at all.

Scans and OCR

A scanned PDF is a photograph of a page and has no text to restructure. With OCR left on auto, the tool recognises that case and runs text recognition first, so the result still has real, selectable, reflowable text. With OCR turned off entirely, a scanned PDF cannot be converted, since there is no text to work with.

Processed on our server, deleted automatically

Reconstructing document structure from raw PDF coordinates needs a real conversion engine, so the file is processed on PulsarKit's server. The uploaded file is deleted as soon as the job finishes. The result stays downloadable for one hour and is wiped from the server within about two hours, whether or not you download it.

Frequently asked questions

Why is the text restructuring only approximate?

A PDF stores where characters sit on the page, not the document's structure, so headings, paragraphs and reading order all have to be guessed by Calibre's heuristics rather than read directly. It works well on straightforward documents and less well on complex layouts.

Which PDFs give the best results?

Text-based, single-column documents: novels, reports, most books. Multi-column layouts, tables and heavy use of headers and footers are the cases where reading order can come out wrong.

What happens with a scan that has no text layer?

With OCR set to auto (the default), this is detected automatically and text recognition runs first, so the result still has real text. Set OCR to never and a scan without text cannot be converted.

What does “auto” columns mean, versus “one” or “two”?

Auto lets Calibre's own layout analysis decide. “One” tells it to treat the whole page as a single reading column regardless of what it looks like, which helps when auto-detection puts a two-column paper's columns in the wrong order. “Two” is for the opposite case, a document Calibre reads as one column when it has two.

Are images kept?

Yes, unless you turn that off. Images are carried through from the PDF into the ePub by default.

How long does the file stay on the server?

The uploaded PDF is deleted as soon as the job finishes. The converted ePub stays downloadable for one hour and is wiped from the server within about two hours, whether or not you download it.