Text & data
Convert PDF to Markdown
Drop in a PDF and get Markdown with headings, paragraphs and lists — not one line per printed line. Convert several at once and take them as a ZIP. The file never leaves this browser.
Try it now — no account neededSave and continue on the desktop
Download or copy your result here first. Open the free desktop to save to Files or use more features.
- Quick notes in a stream, or titled notes
- Save to Files when you continue on the desktop
- Every tool in one place — no new tabs
After you download — continue on the desktop for Files or the full app.
How it works
- Add PDFsChoose one PDF or several — nothing is uploaded.
- Recognize a scanIf a file is a scan, tap Recognize text on its row; text PDFs convert straight away.
- Copy or downloadTap Copy Markdown or Download .md — or Download all as ZIP for several files.
What PDF to Markdown can do
Paragraphs, not printed lines
A PDF has no paragraphs. It has glyphs at coordinates, and everything a reader sees as structure is inferred from where they sit and how big they are. Copy from a PDF reader and you get exactly that: a hard line break wherever the page happened to wrap, a hyphen in the middle of every word that was split at the margin, and a page number in the middle of a sentence. That output is unusable in a document and worse in a prompt. Here the lines are rejoined into the paragraphs they came from, split words are put back together, and running headers and footers — the same title repeated forty times — are dropped.
Headings come from the type size
Nothing in a PDF says "this is a heading". What it says is that one line is set at 18 points among body text set at 10, so that line becomes an H1, the 13-point ones become H3s, and a long line in slightly larger type stays a paragraph — which is the guard that keeps a whole introduction from becoming a title. It is inference, and it is right far more often than it is wrong, but it is worth knowing that it is inference rather than a structure the file was carrying.
Scans are read too
A scan has no text layer — its pages are pictures of letters — so its row offers «Recognize text». Recognition runs in your browser, in the page’s language plus English; it downloads a reading model once (4–5 MB) and uploads nothing. Up to 20 pages at a time. Check names, numbers and dates in what it reads.
What it will not pretend to do
Text and headings, not layout: two-column papers are read as text, and figures, charts and text boxes are not rebuilt. Tables are read by where their text sits: rows of cells in clear columns become a GFM pipe table, while merged cells come through as plain text, because a pipe table has no way to write them. Where it cannot tell, it gives you the text rather than a guessed table.
Related apps
Questions
Is my PDF uploaded anywhere?
No. It is parsed in your browser, on this device. The file is never sent to a server, and we never see it.
Can I convert several files at once?
Yes. Add as many as you like — each becomes its own `.md`, and one button takes them all as a ZIP.
What happens with a scanned PDF?
It has no text layer, so the tool says so and offers «Recognize text». Recognition runs in your browser, in the page’s language plus English; the first time it downloads a reading model (4–5 MB). Check names, numbers and dates afterwards — recognition can misread a character.
Are tables converted?
Yes, when its columns line up. Nothing in a PDF marks a table as one, so it is read from where the text sits: rows of cells aligned in columns become a GFM pipe table — on a scan too. Merged cells and a table pasted in as a picture come through as text. Tables from `.docx` files convert as they are.
Does it work offline?
Yes, once you have converted one PDF here while online: the PDF reader the conversion needs loads with the first file, and your browser keeps it after that. Reading a scan needs one online run as well, for the reading model.