Help Center

Yes. The PDF is parsed by PDF.js running in your browser. There is no upload step. If you open the network tab in your browser's dev tools while converting, you'll see the page itself loads but no request is made carrying your file. After the page loads, you could even turn off your wifi and conversion would still work.

Almost always a scanned PDF. Try selecting a sentence with your mouse in the original PDF. If you can't, the PDF is a photograph of a page with no real text, and PDF2MD has no OCR. Run the file through Preview's "Recognise Text" on Mac, or through a tool like Marker that has OCR built in, then try again.

Tables in PDFs are not really tables — they're text positioned in a grid, with no metadata that says "this is a cell." Every converter has to guess. The most common failure is wrapped cells (a cell whose text runs onto two lines gets read as two rows). If structure matters, my fix is usually to retype that one row by hand. There's a longer write-up on the blog under "Understanding PDF Structure."

There's no hard size cap because processing happens locally — limits depend on your machine's RAM. I've converted 200-page documents fine. Batch is set to 100 files at a time as a soft cap; if you need more, do two batches. Conversion is free and there's no registration.

PDFs that were exported from Word, Pages, LaTeX, or any digital source. If the text is selectable when you open it in any PDF viewer, PDF2MD will handle it. Multi-column academic papers mostly work but occasionally jumble; if that happens, try Marker for that document.

Inline math sometimes survives. Display equations usually don't, because they're rendered as positioned glyphs without semantic structure. If you need math preserved, the source is probably available as LaTeX somewhere (most arXiv papers are), and that's a much better starting point than the PDF.

If the PDF requires a password to open, PDF.js can't read it without the password. Unlock the file in your normal PDF viewer (re-save without the password), then drop the unlocked copy into PDF2MD.

Yes. Drag the folder onto the page, or select multiple files in the file picker. After conversion, hit "Download all as ZIP" to grab everything at once. I do this several times a week with research folders.

Currently no — PDF2MD outputs text-only Markdown. If your PDF has diagrams or screenshots and you need them, you'd need to extract them separately (most PDF viewers let you right-click and save an image). Image export is on my list to add but not done yet.

Once the page has loaded, yes. The PDF parsing engine is shipped with the page itself. You can disconnect your network and keep converting. If you want it to be really offline, install the page as a PWA — most modern browsers will offer this from the address bar.

Recent versions of Chrome, Firefox, Safari, and Edge. Anything from the last two years should be fine. Internet Explorer is not supported and won't be.

Some PDFs include running headers like "Page 12 of 47" on every page. The converter dutifully includes them because, from its point of view, they're just text. Easiest fix: search-and-replace once after converting. I usually grep for "Page \d" and remove.

The repo is at github.com/littleben/pdf2md. Issues and pull requests welcome. If you want to self-host the converter UI on your own infra (e.g. for an internal tool), the project is small enough that this is straightforward.

Either open an issue on GitHub or email [email protected] with the PDF attached if you can share it. I keep a folder of pathological cases and use them to test improvements. Real broken PDFs are the most useful bug reports I get.