A PDF holds formatting you cannot paste into a Markdown editor. Font sizes, bold text, bullet points. This tool reads those clues and rebuilds them as Markdown, inside your browser, without uploading a single byte.
Markdown output
Why Markdown from a PDF is not a one-to-one copy
A PDF stores what each character looks like, not what it means. Every letter has a font, a size, and a position on the page. Nothing in the file says "this is a heading" or "this is a bullet list".
Markdown is the opposite. It is pure structure. A heading is a hash mark. A bullet is a dash. A link is wrapped in brackets. Nothing about how it looks.
The gap between the two is why this conversion requires heuristics. The tool reads visual clues and decides what the structure probably was.
The four clues the tool reads
Font size. Body text has a consistent size. Anything noticeably larger is likely a heading. The ratio tells the level: 1.3x becomes an h3, 1.6x becomes an h2, 2.0x becomes an h1. This matches how most documents are designed.
Font weight. Font names that contain "bold" or "heavy" produce bold Markdown for short paragraphs. Font names that contain "italic" or "oblique" produce italic.
Line prefixes. A line that starts with a bullet character becomes a list item. Same for numbered lines like "1." or "a.". The bullet character is dropped and replaced with a Markdown dash.
Vertical gaps. Consecutive lines inside a paragraph sit close together. A larger vertical gap separates paragraphs. The tool measures the median gap between lines and uses that to decide where paragraphs break.
Toggles for each detection
Every detection is optional. Turn off headings, and every block comes out as plain paragraphs. Turn off lists, and the bullet characters stay in the text as-is.
This matters because detection is a guess, not a certainty. Some PDFs have sections that look like headings but are actually just large pull quotes. Others have short bold statements that should stay as bold, not become headings. If a document converts oddly, the fastest fix is to toggle off the wrong detection and copy again.
Page breaks add an HTML comment between pages. Some Markdown editors render comments as hidden; others show them as text. Turn it off if you do not want page separators in the output.
What converts and what does not
Five structural elements convert reliably.
- Paragraphs with proper spacing
- Headings at three levels
- Bold and italic short paragraphs
- Bullet and numbered lists
- Link targets from PDF annotations
Four elements do not convert and are worth knowing about.
- Tables. PDF has no table structure to read. Table rows come out as separate text lines.
- Images. They are not extracted. Only text is converted.
- Multi-column layouts. Columns interleave the way they do in any text extractor.
- Code blocks. Monospaced paragraphs are detected as text, not fenced code. You can wrap them manually if needed.
The CommonMark specification defines the Markdown features the output targets. Anything the tool produces will paste cleanly into Obsidian, Notion, Bear, GitHub, and any CommonMark-compliant editor.
Where the output goes
Copy the current view to your clipboard, or download a .md file. Both act on whatever is in the text area: the whole document, a single page, or a range you typed.
The PDF loads into browser memory. Markdown is built locally. Nothing is uploaded, logged, or stored on a server. If the source PDF needs trimming first, use Delete PDF Pages before converting. If the PDF is a scan and returns nothing, the text layer is missing - try PDF to Text first to confirm.