Turn a PDF into a Markdown file for your notes app

A PDF stores where each glyph was drawn, not whether a line was a heading or a bullet, so the tool rebuilds structure from clues. It finds the font size that carries most of the text and treats that as body text. Any size noticeably larger becomes a heading, with up to four levels ranked largest to smallest. A short line that is mostly bold, in a size not already ranked as a heading, with no period, comma, colon or semicolon at the end and not followed directly by another bold line, becomes a heading one level below those.

Lines that start with a bullet character, or a number, letter or roman numeral followed by a period or closing parenthesis, become list items. Indented items nest up to three levels deep, judged by distance from the left margin. Ordered lists are renumbered from 1, so a list that started at 7 in the PDF starts at 1.

Wrapped lines are joined into paragraphs and end-of-line hyphenation is undone. Bold and italic runs come through as Markdown emphasis. Text in a monospace font is wrapped in a fenced code block. Tables are detected when three or more consecutive lines break into two or more short cells that line up in columns, and come out as pipe tables. When the typical cell runs past about forty characters, the tool treats the lines as prose in columns and text-heavy tables become paragraphs.

The five checkboxes under "What to look for" default to a sensible state for notes. Keep "Headings" and "Lists" on; they make the file navigable in an outline view. If aligned text keeps turning into stray tables, turn "Tables" off and run again.

"Drop repeated headers and footers" is on by default. It removes lines in the top or bottom twelve percent of the page that repeat on at least sixty percent of pages, and on at least three, with digits ignored so "Page 3 of 40" and "Page 4 of 40" count as the same line. Turn it off if the running head carries something you want, such as a chapter title.

"Mark page breaks" is off by default and inserts a horizontal rule between pages. Turn it on if you cite page numbers or plan to split the file into one note per page.

After you click "Convert to Markdown", the page says "Your file is ready" with a note such as "Found 12 headings, 40 paragraphs, 3 lists." and a preview labeled "Markdown" with a "Copy" button. Obsidian keeps pasted Markdown as Markdown. Notion turns pasted Markdown into its own blocks; if a table arrives as plain text, import the .md file instead of pasting.

For longer documents, use the download button, which carries the file's name, such as "Download handout.md". Drop the file into a folder in your Obsidian vault, or use Notion's import option for Markdown files. The tool converts one PDF at a time, so to turn a long book into several notes, split it into chapters first with Split PDF, then convert each part.

Two-column papers are the main problem. The tool groups text into lines by vertical position across the whole page width, so a line in the left column and the line beside it in the right column are joined into one line, and sentences from both columns run together. No option fixes this. Check the preview, and for affected pages plan to rebuild the paragraphs by hand or take the text from a PDF reader's selection.

Footnotes land in the body text near the bottom of each page as a small paragraph. Headings are missed when a document uses the same size and weight for everything, and a bold lead-in line with no closing punctuation can be promoted to a heading by mistake. Both are a one-line edit.

If you see "No text was found", the PDF is a scan with no text layer. Run it through OCR PDF first, which recognizes English text in your browser and adds a hidden text layer, then convert that file. That hidden text uses one plain font sized to each word, so larger printed headings can still be picked up by size, but nothing is marked bold, so expect to fix some heading levels.