About PDF to Markdown
PDF to MD is useful when a PDF has to go into a wiki, a docs site, a Git repository, Notion or Obsidian, or be pasted into an AI chat as tidy text. Upload the file and you get Markdown with hash headings for titles, dashes and numbers for lists, pipe tables for tables, and ordinary paragraphs in between. Turn on Save the pictures too and you get a ZIP with the Markdown file and an images folder, each picture linked where it sits in the text, ready to drop into Obsidian, Notion or a static site.
Headings are recognized from the type size, so the structure of the document comes through. Running headers, footers and page numbers are left out, and paragraphs are joined back together instead of breaking at every line. Scanned pages are read too: Recognize text (OCR) on Auto-detect reads only the scans and the pages whose fonts come out as gibberish, Always reads every page, and Off none. OCR reads printed English and Chinese. Pages converts just a chapter or a range. The result appears on the page with a Copy button and is also saved as a UTF-8 .md file. Free accounts get OCR on a set number of scanned pages per file, and VIP on every page.
How to use PDF to Markdown
- 1Upload a PDF
Choose or drop the file. Free accounts can convert PDFs up to 50 MB, VIP up to 1 GB.
- 2Set the options and convert
Auto-detect reads scanned pages with OCR by itself. Tick Save the pictures too for the images, set Pages for part of the file, then click PDF to Markdown.
- 3Copy or download
Use Copy to paste the Markdown straight into your editor, or download the .md file or the ZIP with its images.
Why use Cubfile for this
- Real Markdown structure
Headings, bullet lists, numbered lists and tables use standard syntax that GitHub and most editors render, without page numbers, headers or footers.
- Pictures linked in place
With Save the pictures too, each image lands in an images folder and is linked where it sits in the text.
- Scans read page by page
Only scanned pages and gibberish fonts are read with OCR, while pages with real text keep their own.