Extract Text from a PDF

Features

• Extract by page, with an optional page range
• Lines rebuilt from coordinates: similar y is one line, ordered by x
• CJK text is not spread out one character per space
• Optional page markers, so you can follow the original
• Progress shown, so a long document tells you which page it is on
• A per-page report: the characters on each page and exactly which pages have no text, with the page numbers
• Tells scans, all-whitespace pages and header-only pages apart — so you book the right pages for OCR
• Plain text you can copy, or download as .txt
• Nothing is uploaded and no server is involved

How to use

1. Drop a PDF in
2. Give a page range if you only need part of it
3. The text appears below — use Copy or download for long documents
4. A second card is the per-page report: which pages have no text, so you know which to send to OCR
5. For scans, run OCR first

FAQ

Why does a scan produce no text?

Because a scan is one image per page with no extractable text data — the letters are "painted" in pixels rather than "written" in the file. Getting text out of one requires OCR (optical character recognition), a different kind of tool. You can export the pages with PDF-to-image first and hand those to any OCR service. The test is simple: can you select the text in your PDF reader? If yes, it can be extracted; if no, it is a scan.

The paragraphs come out scrambled, or the two columns are mixed together.

That is inherent to multi-column layouts. The rule is geometric: both columns share the same y for a given line and differ only in x, so grouping by y reads them as one line. Single-column documents (contracts, papers, notices) are unaffected. Headers, footers and footnotes also get mixed into the body for the same reason — they have positions but no semantics, and no extraction tool can avoid that; they only differ in how they handle it.

Why is the order different from what I see?

Extraction follows coordinates, and the source order need not match the visual order — in a layout program the heading may have been added after the body (later in the object list, but higher on the page). This tool reorders by y top-to-bottom and x left-to-right, which usually matches what you see, but full-width headings and side notes can still land oddly. Check against the original when exact layout matters.

Is the file uploaded?

No. Extraction runs through pdf.js in the browser (inside a Web Worker), with no network requests at all, and it works offline. Contracts and reports should not pass through a third-party server.