Turn a PDF into clean Markdown with headings, paragraphs and lists rebuilt. Pick the pages you need, see how many tokens it is for ChatGPT or Claude, and download a .md file. Runs in your browser, so the PDF never leaves your device.
Free PDF to Markdown converter that runs in the browser with PDF.js. It rebuilds headings from font sizes, joins wrapped lines into paragraphs, turns bullets into Markdown lists and drops bare page numbers. Choose a page range, then copy or download .md. It shows the word and token count and which models fit the document in one prompt.
Your Markdown shows up here
Headings, paragraphs, bullet lists and page breaks are rebuilt from the PDF's text layer.
Was this tool helpful?
A PDF stores positioned lines of text, not headings or paragraphs. The converter works the structure back out from font sizes and spacing. This is the real output for a short page.
# Onboarding Guide ## First week Your laptop arrives on day one. Sign in with the temporary password from your welcome email and change it right away. - Set up two-factor sign-in - Join the #team channel ### Who to ask Your manager books a 30-minute check-in for Friday.
Exports from Word, Google Docs, LaTeX and most reports work. A progress bar shows each page as it's read.
Type a range like 1-3, 5. Turn heading detection or --- page breaks off if you don't want them.
The file keeps its name with a .md extension. The Ready for AI bar shows whether it fits one prompt.
It depends on whether the PDF has a text layer. Try selecting a word in your PDF reader. If you can, it has one.
| Text PDF | Scanned PDF | |
|---|---|---|
| Made by | Word, Google Docs, LaTeX, export or print to PDF | A scanner, a phone camera, a fax |
| Text layer | Yes, you can select words | No, each page is an image |
| This converter | Headings, paragraphs and lists | Finds no text and says so |
| Better option | This tool | OCR or a vision model |
Count its tokens exactly, and pick a model that reads long documents well for the price. For scans, pick one that reads images.
Drop a PDF into the converter or click Upload. It reads the text layer page by page, rebuilds headings, paragraphs and bullet lists, and shows the Markdown as soon as it's done. Pick a page range if you only need part of the file, then copy the result or download it as a .md file.
No. A scanned PDF is a picture of each page with no text layer, and this tool extracts text without OCR. It tells you when most pages have no text. For scans, use an OCR tool or a vision model that reads images; the Best model for OCR page ranks them on intelligence, cost per 1,000 pages and speed.
It finds the most common font size in the document, which is the body text, and compares every line to it. Lines at least 1.6 times larger become #, 1.3 times ##, and slightly larger or short bold lines ###. Headings that wrap onto two lines stay one heading. Turn off Detect headings to get plain paragraphs.
Yes. Type a page range such as 1-3, 5 or 10- in the Pages field and the Markdown updates at once, without reading the file again. Leave the field empty for every page. The footer shows how many pages, words and tokens are in the current selection.
No. The file is read in your browser by PDF.js, Mozilla's open-source PDF library, and the Markdown is built on your device. Nothing is sent to Dervity or anyone else, so it's safe for contracts, reports and other private documents. Files up to 100 MB work, and big ones just take longer.
Images are skipped, because only text is extracted. Table cells come through as plain text in reading order, without Markdown table pipes, so check tables by hand. Headers and footers that are only a page number are removed, and words split with a hyphen at the end of a line are joined again.
Markdown keeps the headings and lists that tell a model how the document is organized, in far fewer tokens than HTML. The Ready for AI bar shows an estimate of the token count and which popular models can take the whole document in one prompt, based on their context windows.
No. PDF.js can't open a PDF that needs a password, and the converter tells you when that's the problem. Open the file in a PDF reader with the password, save or print a copy without protection, then drop that copy into the converter.
Everything from model picks to CSV converters, with no account needed.