1. Font-Size Heading Hierarchy
Instead of dumping flat lines of text, the extractor builds a font-size distribution across your PDF to find the body text size, then promotes larger section titles into # H1, ## H2, and ### H3 Markdown headings.
THE TOOLBOX / TEXT & MARKDOWN CONVERTERS
Turn any digital PDF into clean, LLM-ready Markdown or plain text with automatic heading detection, list reconstruction, de-hyphenation, repeated header stripping, and live GFM preview—100% locally.
Detects headings, bullet lists, numbered lists, and paragraph breaks automatically—100% locally in your browser with LLM token count.
PDF TO MARKDOWN WORKFLOW
Instead of dumping flat lines of text, the extractor builds a font-size distribution across your PDF to find the body text size, then promotes larger section titles into # H1, ## H2, and ### H3 Markdown headings.
Academic papers and reports wrap words with trailing hyphens and repeat running headers on every page. Enable Join hyphenated line wraps and Strip repeated headers/footers for clean, continuous prose.
Edit the extracted Markdown directly in the left pane while verifying the rendered GitHub-Flavored Markdown on the right, alongside live word, character, and estimated LLM token counts.
Working with Markdown or PDFs? Try our Markdown to PDF Preview, Markdown Table Generator, PDF to PNG Converter, or Image to Text (OCR).
| PDF Element | Markdown Output | Detection Heuristic |
|---|---|---|
| Document & Section Titles | # H1, ## H2, ### H3 | Font size 1.12× to 1.55×+ larger than body baseline |
| Bullet Points (•, ●, –, —) | - List item | Leading bullet symbol normalized to standard hyphen |
| Numbered Steps | 1. Step | Leading numeric prefix preserved in sequence |
| Wrapped Paragraphs | Continuous paragraph blocks | Vertical line spacing & trailing hyphen joining |
| Multi-Page Boundaries | <!-- Page N --> + --- | Optional toggle for chunking & citation tracking |
It inspects the PDF text layer’s font size histogram and line coordinates locally in your browser. Dominant text sets the body size baseline, significantly larger lines become # H1, ## H2, and ### H3 headings, and bullet/numbered prefixes convert cleanly into Markdown lists.
Raw PDF copy-paste often introduces broken mid-sentence line wraps, repeated page headers, and lost section hierarchy. Clean Markdown uses fewer LLM tokens, preserves document structure, and pastes directly into AI prompts, Obsidian, Notion, and GitHub.
No. Extraction runs 100% client-side using a local PDF engine in your browser tab. Confidential contracts, financial statements, research papers, and internal specs never leave your device.
Because this tool reads the embedded digital text layer of a PDF, image-only scans will return empty text. For scanned documents, render pages to PNG and run our local OCR tool (Image to Text) to extract text from scans.
Yes. Enter any page selection such as "1-5, 12, 20-24" in the Page Range field and click Re-Extract to pull only the sections you need.
It uses the standard ~4 characters per token approximation for English Markdown, giving you an instant sense of how much context window your extracted document will consume.