How to convert a PDF to Markdown
Drop the PDF on the page or pick it with the button. The conversion starts at once and shows the original page on the left and the Markdown on the right, so you can check a heading or a table against the page it comes from. Type a range in Pages to keep only part of the document, choose whether the pictures go into a ZIP or are left out, then copy the Markdown or download it. With several PDFs, each gets a tab, and Download all gives a ZIP with one folder per file.
What comes out as Markdown
Headings become # to ### according to their size, paragraphs are joined again even when a page cuts them, bulleted and numbered lists keep their levels, and bold, italic, strike-through, superscript and links stay on the words they cover. Text set in a monospace font becomes a code block. Tables drawn with lines, and columns aligned on tab stops, become GitHub tables. Page numbers, running titles and other lines repeated on every page are left out unless you turn that off.
Scanned PDFs and password-protected PDFs
A scanned page is a picture: it holds no text. With Read scanned pages on, those pages are read by text recognition in your browser, in the language you pick, and their lines and paragraphs are rebuilt like any other page. A PDF that asks for a password to open shows a password field: the file is opened with it in your browser, and the password is never sent anywhere. A PDF that only restricts copying or printing is converted directly.
PDF tables to Markdown tables
Two kinds of tables are recognised: tables drawn with borders, including merged cells, and borderless tables whose columns line up on the same positions, like an invoice or a table of contents. Both become GitHub-flavoured Markdown tables, the first row used as the header.
Markdown tables cannot merge cells: a merged cell keeps its text in its first column and leaves the others empty. Several lines in a cell are joined with a line break.
Extract the pictures of a PDF with the Markdown
With Pictures set to In a ZIP, every photo, logo or chart embedded in the PDF is saved in an images folder (images/page-3-1.jpg for the first picture of page 3) and the Markdown links to it where it was on the page. A picture repeated on every page, such as a logo, is saved once.
Unzip the archive and open the .md in Obsidian, VS Code or any Markdown editor: the pictures show. Charts drawn as vector shapes are not pictures and are not extracted.
Convert a scanned PDF to Markdown with OCR
Pages without text but covered by a picture are recognised as scans. Text recognition (Tesseract) reads them in the language chosen in the list: English, French, Spanish, German, Italian, Portuguese, Japanese, Chinese, Korean, Arabic, Russian and more.
Recognition takes a few seconds per page and depends on the quality of the scan: check names and numbers against the original page shown beside the result.
Markdown for Obsidian, Notion and GitHub
Turn on Front matter to start the file with a YAML block holding the title and the pages converted: Obsidian shows them as properties, and static site generators read them. Unzip the ZIP into your vault so the images folder sits next to the note.
Notion imports .md files (Import › Markdown), GitHub renders them in a repository or a wiki. A marker where each page starts adds a rule and an HTML comment with the page number, invisible once rendered, to find a passage in the original PDF.
Give a PDF to ChatGPT or Claude as Markdown
Language models read Markdown well: headings and tables tell them how the document is organised, which a flat text extraction loses. Converting first also removes repeated headers, footers and page numbers that waste space in the prompt.
Paste the Markdown in the chat, or attach the .md file. Keep only the pages you need with the Pages field to stay within the length a chat accepts.
PDF to Markdown, to text or to Word: which one?
Pick Markdown when the text goes into notes, documentation, a static site or an AI chat: it keeps the outline (headings, lists, tables) in a light text file that any editor opens and that version control can compare line by line.
Pick plain text when you only need the words, for a quick copy or a word count. Pick Word when the document must keep its exact look, its fonts and its page layout to be edited and printed again.
Check the Markdown against the original
The original page is shown beside the Markdown: use the arrows to move through the pages and compare. The counters above say how many headings, tables and pictures were found, and how many scanned pages were read, so a missed table shows at a glance.
Rendered view shows the Markdown as a reader will see it, with its tables and pictures; Markdown source shows the text you copy. Changing an option converts the file again at once, without counting a new use.
What a PDF can and cannot give back
Layouts made of free text boxes (posters, magazines) come out in reading order, column by column, but their exact placement is lost: Markdown has no positions. Formulas come out as the characters they are drawn with, not as LaTeX.
Links are kept when the PDF stores them as links; a web address written as plain text stays plain text, which most Markdown viewers make clickable anyway.