PDF to Markdown, without uploading the document
Markdown has quietly become the format that machines and humans agree on. It is what static site generators build from, what wikis and note apps store, and — increasingly the reason people want this conversion — what language models read most reliably. A PDF pasted into a chat box arrives as a jumble; the same document as Markdown arrives with its structure intact.
The catch is that the documents worth feeding to an AI assistant are usually internal: a contract, a specification, a research paper under licence, a board pack. Uploading those to a conversion service to prepare them for analysis defeats the purpose. This tool runs the extraction in your browser, so the file never leaves your device.
What the converter reconstructs
A PDF does not store paragraphs, headings or lists. It stores glyphs at coordinates — which is why copying text out of one so often produces a mess. Rebuilding structure means inferring it from geometry.
Lines are grouped from runs sharing a baseline, with the tolerance scaled to glyph height so a large heading is not split in two. Paragraphs are separated where the vertical gap grows beyond normal line spacing. Headings are lines set larger than the document's body size, with the size ladder mapped onto heading levels. Lists are recognised from leading bullet glyphs and numbering, and re-emitted as Markdown list syntax. Characters that would otherwise be read as Markdown markup are escaped, so a literal asterisk in the source stays an asterisk.
Where it stops
Tables come through as plain lines, because most PDFs give no indication that a group of text runs is a grid. Multi-column layouts — academic papers especially — are read in the order the file stores them, which is usually column by column but is not guaranteed. Images and equations are not extracted. And a scanned document has no text layer at all, so there is nothing to convert; the tool reports that rather than returning an empty file.
These are limits of how most PDFs are written as much as of this implementation; a PDF tagged for accessibility can carry more structure, which this tool does not read. Knowing where the output needs checking is more useful than a promise it will be perfect.
Related tools
For an unstructured dump, PDF to Text is simpler. For an editable document, use PDF to Word. For a styled web page, PDF to HTML. To count what you extracted, try the Word Counter.