Free tool · nothing leaves your browser

PDF to Markdown converter

Choose a PDF and get markdown back. The conversion happens inside this page — your document is never uploaded, to us or to anybody else.

Drop a PDF here Up to 60 MB. The file is read inside this page. Disconnect your network once the page has loaded and it still works.

Your file never leaves your browser. There is no upload and no server — the reading and the conversion both happen here, in your own computer's memory. Most free converters take a copy of your document first.

Nothing converted yet Choose a PDF above. The markdown appears here, with a preview beside it.

How to use it

  1. Choose a PDF, or drop one on the box. The reading library is fetched the first time you pick a file, so the page stays light until you use it.
  2. Watch it work through the pages. The counter says which page it is on, so a slow document never looks like a broken one.
  3. Copy it, or download the .md file. The preview beside the markdown is for checking the headings and lists landed where you expected.

What comes out

A page with a heading, a wrapped paragraph carrying bold and italic, and a short bulleted list — run through the same code your browser will use:

# Field Notes

A **paragraph** in a PDF is stored as loose runs of letters at fixed positions, one line at a time, with no idea that it is a *paragraph* at all.

- Wrapped lines are joined back up.
- Real breaks are kept.

The wrapped lines are joined back into one paragraph, the emphasis survives, and the bullet characters become real markdown list items. That is the whole job: a PDF stores where every letter sits and nothing at all about which letters form a heading or a paragraph.

Why "nothing is uploaded" matters here

Most free PDF converters upload your document to a server. Your file goes to somebody else's computer, is converted there, and comes back. Whatever the privacy policy says, a copy of your contract, your medical letter or your unpublished manuscript has left your control.

This page does not, and you need not take that on trust. Open your browser's network panel and convert something: the only things fetched are the conversion libraries. Turn your network off after the page loads and it still works. There is no upload button because there is nowhere to upload to.

Text PDFs and scanned PDFs are two different jobs

A text PDF — anything saved out of a word processor, a layout program or a web page — holds the actual letters, each with a position and a font. Converting one is fast and accurate, and the font is what tells us a line is a heading or a phrase is bold.

A scanned PDF is a picture of a page. There are no letters in it, only dots. Reading one needs optical character recognition: software that looks at the picture and works out which letters it can see. This page offers that and is honest about the cost — about 14 MB of extra download the first time, a few seconds to a minute per page, and some words read wrong.

You are not asked to guess which you have: a page that yields almost no text is named, with recognition offered, rather than handed to you as an empty box.

What it can and cannot work out

It handles

  • Headings, judged against the document's own body text rather than a fixed size — the only way to find them in a document that sets every heading in the body's own size, in bold.
  • Paragraphs: wrapped lines joined back up, real breaks kept, sentences carried across a page break, and words split by an end-of-line hyphen put back together.
  • Bulleted and numbered lists, including nesting, from where the markers sit.
  • Bold and italic, taken from the font each run of letters is set in.
  • Links, where the PDF carries them as real link areas.
  • Page numbers and running headers, dropped rather than scattered through the text.

It warns instead of guessing

A converter that quietly interleaves two columns hands you something that looks converted and is nonsense. So where it cannot be sure, it names the page:

  • Two-column pages are read down one column and then the other, and named so you can check.
  • Tables are reported and left as plain lines. They are not rebuilt: a table with its cells in the wrong order is worse than no table.
  • Scanned pages are named, and recognition is offered for them.
  • Lists lettered a, b, c (or i, ii, iii) become ordinary numbered lists. Markdown has no lettered list, so the letters themselves are lost.
  • A document with no headings is told to you as exactly that, rather than handed back as a flat wall of text that looks like it worked.

It cannot do

Images are not pulled out. Footnotes are not linked back to their marks. Equations come out as the letters they are made of. Documents in scripts that need a character map — Chinese, Japanese and Korean especially — may extract poorly, because those maps are not part of what this page loads. Mathematical and heavily designed documents will need a hand afterwards.

Common questions

Is my file really not uploaded?

Really. The conversion runs in your browser through PDF.js, the same engine Firefox uses to show PDFs. Your document goes from your disk into your browser's memory and no further. Disconnect your network after the page loads and try it.

Is it free? Is there a limit?

Free, no account, no watermark, no cap on how many documents you convert. The only limit is your own computer: the file has to fit in the browser's memory, so the page stops at 60 MB.

How good is the text recognition?

Good enough to read, not good enough to publish unchecked. On a sharp scan we measured about 99 words in 100 coming through exactly right; on a soft, low-resolution one, about 4 in 5. Headings are found only where they are set in capitals or in much larger letters, and bold and italic are gone entirely — a picture of a page carries no fonts, so there is nothing left to read them from. Proofread the result.

Why is the text recognition so big?

It ships a whole recognition engine and an English language model into your browser rather than sending your page away to be read. That is the trade: about 14 MB and some patience, for a picture of your document that never leaves your computer. It loads only if you ask.

My headings came out as ordinary paragraphs.

Then the PDF gave nothing to tell them apart — same size, same weight, same position. Some design tools export a heading that is only visually a heading, and a few PDFs carry no font information at all. When that happens the page says so under the result rather than guessing.

Can I get a Word file, or HTML?

Markdown only. It is plain text, so it opens in anything, and most writing tools import it.

Something complicated did not convert well.

Say what it was — complex-layout reconstruction can be turned on for an account on request. Send us the details and we will look at the document.

While you are here

The rest of Final Boss Games is daily puzzles and small browser games. The other tools here — a word search maker, a cryptogram maker — follow this one's rule: free, no sign-up, and nothing you type or open leaves your browser.