PDF → Markdown
Turn PDF documents into clean, editable Markdown. Private, free, and processed in your browser.
Choose one file. Conversion starts automatically.
Text PDFs use their existing text. English scans use local OCR only when needed.
A simpler way out of a PDF
Bring the structure into Markdown, review it beside the original, then keep the version you need.
One PDF · up to 64 MB · English OCR when needed
Copies and downloads use your current edits.
Useful beyond the download
A PDF fixes text to a page. Markdown describes headings, paragraphs, lists, and links using a few readable characters. This tool rebuilds that structure where the PDF provides enough evidence, giving you a .md file you can edit in a notes app or text editor.
Reuse a handout in your notes, bring a report into documentation, or keep a searchable copy in a knowledge base. You can read the rendered preview without learning Markdown, then check the content against the original PDF.
PDF.js reads the document’s text, fonts, coordinates, and links. The converter groups lines into paragraphs, checks likely headings and tables, and orders common two-column pages by column. Uncertain structure stays as text instead of being presented as a verified table.
The download is plain UTF-8 text, suitable for systems that accept Markdown. The converter itself does not use an AI service. If you later add the file to another service, that service’s privacy rules apply. Check reading order and important numbers before relying on the conversion.
Pages without readable text are rendered into images and read locally using English OCR, or optical character recognition. The OCR program and English model load only when needed from HandyPebble’s own paths. Clear, upright English scans work best; handwriting, other scan languages, and complex tables can need manual correction.
Common headings, lists, emphasis, verified links, and regular tables can carry over. Markdown is not a visual replica: photos, charts, precise positioning, form behavior, and attachments are not included. Fonts with missing character information and unusual layouts may need OCR or editing.
Clear answers
No. The PDF, its text, page images, OCR results, and your edits stay in this browser. There is no document upload or AI API. The site still makes ordinary requests for its static files and may generate website analytics traffic; see our privacy policy.
No document data is saved to browser storage. Refreshing, closing, or clearing the page removes the current document and edits. Copy or download your work first. The browser can cache the static conversion program and OCR language files, not your document content.
No. A PDF often stores drawing instructions rather than paragraph or table meaning. The converter uses conservative layout checks and flags uncertain structure. Compare multi-column pages, tables, captions, and important numbers with the original.
Yes. Open Advanced options and enter a range such as 1-3, 5, 8-. Overlapping ranges are deduplicated and pages follow their original order. Select Convert again to apply the range. Preserve page breaks adds a Markdown comment with each source page number.
Existing PDF text can include Unicode, including mixed English and Chinese text. Local scan recognition supports English only in this version. Text from unusual fonts, vertical scripts, or damaged text layers can need a reading-order check or manual editing.
Yes, when PDF.js supports its encryption and you know the password. The page asks for it locally. The password is not saved or sent to a server. Unsupported encryption or a damaged file can still prevent opening.
The file limit is 64 MB and the document limit is 1,000 pages. Convert up to 200 text pages or 20 English OCR pages at once. Text fragments, output length, and page image sizes also have memory limits. Use a smaller range or PDF Toolbox to split a document when needed.
No. The source PDF is unchanged. Your edits affect only the Markdown result, its preview, and the file or text you copy. Raw HTML and scripts do not run in the preview, and external images are not loaded.