Structure, not just words — headings, lists and paragraphs, in your browser.
Processed on your device
Your files are processed in your browser and never uploaded.
Can you convert a PDF to Markdown without uploading it?
Yes. A PDF stores glyphs at coordinates and nothing about structure, so the headings and lists have to be inferred — from font sizes, weight, indentation and the gaps between lines. That inference runs in the browser, so the file stays on your device. Table structure and multi-column reading order are not recovered.
About PDF to Markdown
Plain text extraction gives you the words; this gives you the shape of the document. A PDF stores glyphs at coordinates and records nothing about what is a heading or a list, so the structure has to be inferred — from font sizes, from weight, from indentation, and from the gaps between lines. That inference runs right here in the page, on your own machine: headings become #/##, bullets and numbered lists become list items, wrapped lines are rejoined into paragraphs, words hyphenated across a line break are repaired, and repeated running headers and page numbers are dropped. It is built for getting a document into an AI tool, a wiki or a static-site generator, where the structure is the whole point — and for doing that with a document you would rather not hand to someone else's server.
How to use PDF to Markdown
1
Add a PDF
Born-digital documents work best: reports, papers, letters, documentation. A scan has no text layer to read, so put it through the OCR tool first.
2
Pick a format
Markdown keeps the structure. Plain text gives you the words in reading order with nothing added.
3
Convert
The conversion happens in this page. Preview the result, then download it — nothing was uploaded, so there is nothing to delete afterwards.
Frequently asked questions
Is my file uploaded?
No. The PDF is read and converted entirely in this page, on your own processor. The simplest way to check is to disconnect from the network and convert a document anyway: it still works, because there is no server involved.
How accurate is the structure?
It is an inference, not a parse. A document with consistent heading sizes — a report, a paper, a letter — converts cleanly. A magazine layout, a form, or a page of dense columns will not. Nothing is invented: where the layout gives no signal, the text comes through as ordinary paragraphs.
What happens to tables?
Their structure is not recovered. The cells arrive as lines of text in reading order. A PDF records a table as unrelated text runs and some ruled lines, and rebuilding the grid from that needs a heavier engine than a browser can carry — so rather than guess at a table and get it wrong, the text is kept and the grid is not claimed.
How is this different from PDF to Text?
PDF to Text returns the raw words, line by line. This adds the structure on top: heading levels, lists, wrapped lines rejoined into paragraphs, hyphenation repaired, running headers and page numbers removed. Both run in your browser and neither uploads anything.
Why are the repeated page headers missing from the output?
On purpose. A line that appears in the top or bottom margin of most pages is a running header, a footer or a page number, and repeating it through the middle of the markdown is noise. It takes at least three pages before a repeat counts as a pattern, so a short document keeps everything.
Can it do HTML or JSON?
Not on this site. Those formats come from an optional self-hosted structure engine (see server/README.md in the repository), which also reconstructs tables. It is off here, because switching it on would mean uploading your file — and that is not a trade this site makes by default.