Upload a PDF with a real text layer. The tool reads the positioned text, infers heading levels from relative font sizes, converts bullet and numbered lists, and rebuilds paragraphs. Copy the Markdown or download it as a .md file.
Language models and RAG pipelines work far better on structured Markdown than on raw PDF text, because headings and lists preserve the document's hierarchy. Markdown also drops straight into static site generators and note apps like Obsidian.
Each line's font size is compared against the median font size of the whole document. Lines that are noticeably larger become H1, H2 or H3. This is a heuristic, so unusual layouts may need a quick manual pass.
No. A scan contains pictures of words, not text, so there is nothing to extract. Run the scan through OCR first.
Tables are emitted as plain text lines. For real tabular output, use the PDF Table Extractor to get CSV.