PDF to Markdown

How heading detection works

pdf.js exposes the font size of every text run. The tool clusters sizes:

1. Find the most common (modal) size — that's body text
2. Sizes ~1.4× larger → H2, ~1.8× → H1, ~1.2× → H3
3. Anything below body size → footnotes / small text

This works well for documents with consistent typographic hierarchy. Magazines, decorative layouts, and PDFs exported from inconsistent sources may need cleanup.

What's preserved

• Paragraph breaks
• Heading hierarchy (H1-H3)
• Inline bold (if it's encoded in font name)
• Bullet-list indentation

What's not

• Images (replaced with [image])
• Complex tables (becomes flat text)
• Two-column layouts (text may interleave)
• Footnotes (mixed into body text)

Privacy

The PDF is parsed in your browser via pdf.js. Nothing is uploaded.