PDF to Markdown

No page gave up a stored character other than white space, which is typical of a scan, a phone photo, or a file whose lettering was turned into drawn outlines. OCR PDF can add a text layer: leave its Output on "Searchable PDF" and select every language printed on the pages, then convert the PDF it returns here. OCR PDF runs on pdf.xyz's secure server, so unlike this tool it uploads your file.

Headings, partly. Besides the # title made from the file name and, in a multi-page PDF, the ## Page headings, short lines set at least 20% larger than the body text become Markdown headings, the larger sizes ranked higher. A heading set less than 20% larger, longer than 80 characters, or ending in a full stop, comma or semicolon stays ordinary text. The table grid, bold and italic, link targets and images are lost; table cells come through as words.

A line becomes a list item when it opens with -, * or + and a space, with digits followed by "." or ")" and a space, or with a bullet character such as •, ▪, ● or ◦, including the Symbol and Wingdings bullets some PDFs store in place of •. Every unordered marker becomes "- " and numbers are kept. The rule reads characters, not layout. Bullets drawn as small shapes rather than stored as characters leave nothing to read, so those lines can run into the paragraph, while a line such as "2023. Revenue grew" is split out as if it were numbered.

Words are not rewritten, and they come out in reading order where the order the PDF stores them in is out of step with the page: a title or heading stored after the text under it, a running header stored last, and a footer or page number stored first are each moved to where they read, while a page stored a column at a time comes out a column at a time. Runs of spaces shrink to one. A word hyphenated at a line end is rejoined, so "exam-" and "ple" become "example"; a compound broken at its own hyphen, like "well-" and "known", loses the hyphen too, while one followed by a capital or a digit keeps it, as in "Jean-Paul". A paragraph that would otherwise read as Markdown syntax (one to six # signs and a space, a rule such as --- or ***, a code fence, a > quote or a | table row) gets a backslash so a Markdown viewer shows it as typed, and a < or & that would open an HTML tag or entity is escaped; other asterisks, underscores and backticks pass through and may be formatted by a Markdown viewer.

Both read the text with the same in-browser extraction. PDF to Text writes a .txt file that keeps the PDF's line breaks and adds no markers; PDF to Markdown joins those lines into paragraphs, turns larger type into headings, splits out list lines and adds the title and page headings.

Copy or printing restrictions are no obstacle: a PDF that opens without a password is converted like any other file. One that needs a password to open is refused as it opens, with a message saying it is password-protected; Unlock PDF can remove that password if you enter it, but that tool uploads the PDF to pdf.xyz's secure server.

Not fully. The page and its processing engine load over the network, and parts of the engine are fetched the first time a step needs them. The processing itself runs on your device. Chinese, Japanese or Korean text whose font is not embedded in the PDF needs a character map, downloaded from our site the first time it is needed, and that step needs a connection.

Up to 100 MB per file.

Partly. A PDF stores text as positioned lines, not as headings or tables, so the converter infers structure from size and position. The file's name becomes the title, each page gets a "Page N" heading when there is more than one, short lines set at least 20% larger than the body text become headings, lines are joined into paragraphs, and lines that begin with a bullet (-, *, + or a character such as •) or a number stay list items. Bold, italics, links and table layout are not carried over — table rows come through as plain text — and images are not exported. For tables, Extract Tables detects them by their layout and exports CSV and Excel files.

Not on its own. PDF to Markdown reads the text a PDF already contains, and a scanned page is a picture with no text inside it, so there is nothing to read. Run the scan through OCR PDF first, then use the result here.