PDF to Markdown Converter

A PDF does not store a document the way a word processor does. It stores instructions for painting glyphs at coordinates on a page. There is no record that a line is a heading, that three lines belong to the same list, or that a block of text is a table — that structure is something a reader infers from size, position and spacing. Converting a PDF to Markdown means rebuilding those inferences.

1. Read the text layer. The file is opened with Mozilla's pdf.js inside your browser. Every text fragment comes back with its position, width and font, which is the raw material for everything else.
2. Rebuild lines and columns. Fragments on the same baseline are joined into lines, and wide horizontal gaps within a line are treated as cell boundaries — the signal used later to spot tables.
3. Classify each line. Text larger than the document's body size becomes a heading, with the largest size mapped to #, the next largest to ## and so on. Lines starting with a bullet glyph or a number become list items. Lines set in a monospaced font become fenced code. Running headers and page numbers that repeat in the margins across pages are dropped.
4. Emit Markdown. Wrapped lines are rejoined into paragraphs, words hyphenated across a line break are repaired, link annotations become Markdown links, and characters that would be read as Markdown syntax are escaped.

Because the inference runs on geometry, the output quality tracks how conventionally the PDF was laid out. A report exported from Word, Google Docs, LaTeX or a static site generator converts well. A magazine spread does not.

Element	Markdown output	Reliability
Headings	# to ######	Good when heading sizes differ from body text
Paragraphs	Blank-line separated, wrapped lines rejoined	Good
Bulleted lists	- item, nested by indentation	Good
Numbered lists	1. item, original numbers kept	Good
Links	[text](url) from PDF annotations	Good where the PDF stores real link annotations
Bare URLs	<https://…>	Good
Tables	GitHub-flavored pipe tables	Fair — simple grids only
Code blocks	Fenced with ```	Fair — depends on a monospaced font being used
Bold and italic	Not emitted	Not supported — see limitations
Images	Not extracted	Not supported

Unicode is preserved throughout, including non-Latin scripts, accented characters, mathematical symbols and emoji, provided the PDF embeds the correct character mapping.

Given a PDF page whose first line is set in 24pt, a subheading in 18pt, body copy in 11pt, three bulleted lines, and a two-column grid of figures, the converter produces:

# Quarterly Revenue Report

## Regional performance

Revenue grew across all three regions in the period, with the
strongest contribution from the northern territory.

- North: ahead of plan
- Central: on plan
- South: behind plan

| Region | Revenue |
| --- | --- |
| North | 412,000 |
| Central | 388,000 |
| South | 194,000 |

Full methodology is available at <https://example.com/methodology>.

Note what happened: the 24pt line became an H1 and the 18pt line an H2, not because of any absolute size threshold but because they are the two largest sizes above body text. The wrapped sentence was rejoined into one paragraph. The figure grid became a table because both rows shared the same column edges. The footer URL became an autolink. The page number that sat at the bottom of both pages was removed.

There is no upload. The PDF is read through the browser's File API, decoded by a WebAssembly-free JavaScript parser running in a web worker on this same origin, and converted to Markdown in the page. No request carrying your file contents is ever made, because this site is a static page with no server-side component to receive one.

Two ways to verify that yourself: open your browser's network panel and watch it while you convert — you will see the parser being fetched and nothing being sent — or load this page, disconnect from the internet, and convert a file anyway. It works.

The practical consequence is that confidential material — contracts, medical records, unpublished drafts, anything under NDA — can be converted without it crossing a network boundary. It also means the size limits below exist: your own device is doing the work.

No PDF-to-Markdown conversion is exact, and this one does not claim to be. Expect to edit the output — that is why it is presented in an editable box rather than as a finished download.

• Scanned documents produce nothing. A scan is an image. There is no text to extract, and you will get an empty result with a warning. Run the file through OCR first.
• Bold and italic are not detected. Emphasis in a PDF is usually a separate embedded font rather than a style flag, and inferring it reliably across arbitrary font naming conventions produces more false positives than it is worth. Emphasis is lost.
• Multi-column layouts interleave. Academic papers and newsletters set in two or three columns are read in visual line order, so text from adjacent columns can alternate. Single-column documents are what this handles well.
• Complex tables drift. Merged cells, cells wrapping onto several lines, and layouts using invisible tables purely for positioning will come out misaligned or be missed entirely.
• Images are not extracted. Figures, charts and logos are skipped; only text is converted.
• Footnotes lose their anchors. Footnote text is usually captured, but the superscript reference that pointed to it is not linked.
• Size limits apply. 25 MB and 500 pages, because the work happens on your device.

If your source document still exists in its original format, converting from that will always beat converting from its PDF export. Reach for this tool when the PDF is all you have.