PDF → EPUB
Extract the text of a PDF and rebuild it as a reflowable ebook that adapts to your screen, instead of forcing you to pinch and pan a fixed page.
Drop your PDF
or click to choose — the only ceiling is your own memory, none on our side
Reading a PDF on a phone is miserable for one structural reason: a PDF records where each character sits on a fixed sheet of paper, so shrinking the sheet to fit a small screen shrinks the type with it. An EPUB stores the text and lets your reader lay it out, which is why the same book is comfortable on a phone, a tablet and an e-ink reader.
Getting from one to the other is inference, not translation. A PDF contains no statement that a run of characters is a paragraph, or that a line in larger type is a chapter title — those have to be worked out from position and size. This tool groups characters into lines by their vertical position, joins lines into paragraphs where the spacing says the paragraph continues, and treats noticeably larger lines as headings. On a novel or a report that works well. On a multi-column academic paper it does not, and the page says so rather than handing you rubble.
The extraction is done by pdf.js, Mozilla's PDF engine — the same one built into Firefox — running in this tab. Your file is read by your own browser and the EPUB is written back into memory on your machine. Nothing is uploaded, which for an unpublished manuscript or a document you would rather not hand to a stranger's server is the entire point.
Format by format
| What it does | EPUB | |
|---|---|---|
| Layout | Fixed pages at a stated size | Reflows to any screen |
| Type size | Baked in at creation | The reader chooses, any time |
| On a phone | Pinch, pan, squint | Comfortable at any size |
| Structure | Implied by position and type size | Stated explicitly: chapters, headings, a table of contents |
| Scanned pages | Common, and they are just images | Impossible — there must be real text |
| Best for | Printing, forms, archiving a layout | Reading a book on a screen |
What actually changes
PDF stores where every character sits on a fixed sheet. Nothing in it says “this is a paragraph” or “this is a chapter” — those are inferred from geometry, and the inference is good rather than perfect. Check the result before you delete the original.
What survives
What does not
How it works
It is opened by this page with Mozilla's pdf.js, on your own machine.
Lines are grouped by position, then joined into paragraphs where the spacing says a paragraph continues.
With chapters, a table of contents, and text that reflows to any screen.
FAQ
Your file is a photograph of pages — a scan. The characters are pixels, not text, so there is nothing to extract. Reading them needs OCR, which is far too slow to run in a browser tab for a whole book: expect tens of seconds per page. A PDF exported from Word, InDesign or LaTeX converts fine.
Column detection is the hardest part of this conversion and this tool does not attempt it yet, so the two columns interleave. Academic papers and magazines are the weak case; novels, reports and manuals are the strong one. If your paper has a single-column preprint version, convert that instead.
Not yet — this is a text extraction, so illustrated books lose their pictures. The tool reports the page count and the number of chapters it found so you can judge whether the result is usable before you rely on it.
No. pdf.js runs in your browser and the file never leaves it. You can disconnect from the internet once the page has loaded and the conversion still works, which is the simplest way to prove it to yourself.
Not if it needs a password to open. If the PDF merely restricts printing or copying but opens without a password, it will usually convert.
Very good on a straightforward book: a novel or a report comes through with its paragraphs and chapters intact. Mediocre on anything designed as a layout — multi-column text, tables, sidebars, formulae. It is a text rescue, not a facsimile.
Read it, or run it through a translator. Since it is now a standard EPUB, every other tool on this site accepts it — and DualBook can translate the whole book and hand back a bilingual edition.
Once it is an EPUB, DualBook can translate the whole book and hand it back as a bilingual edition.
