PDF to text

An empty extraction almost always means the PDF is a scan: the pages are photographs of a document and there is no text layer for pdf.js to read. The tool says so directly rather than handing you a blank file, and points at OCR PDF, which runs character recognition over the page images and produces a searchable PDF you can bring back here. The other cause is a PDF whose fonts are embedded as subsets with broken or missing character maps, common in output from old drafting and typesetting software; the page looks like text, but every glyph maps to a meaningless code point, so you get garbage instead of nothing. Character recognition is the fix in that case too, because it reads pixels rather than trusting the encoding.

Layout is not reconstructed — the output is reading order, one line per visual line, with a blank line between pages. Two-column pages interleave badly, because pdf.js reports runs roughly in the order the page paints them rather than in the order a human reads them, and a table collapses into a stream of cell values with no column alignment. That is a deliberate trade: reconstructing geometry accurately is a different problem, and PDFChamp solves it in different tools. If your PDF is mostly tabular, PDF to CSV and PDF to Excel cluster the same text runs by x and y position into real rows and columns. For prose, headings and single-column reports the line reconstruction here is usually clean enough to use as-is.

Copying from a reader gives you whatever that reader's selection model produced, usually with hard line breaks stripped or duplicated and page furniture mixed in, and it becomes painful past a few pages. This tool walks every page programmatically, keeps one output line per visual line by combining pdf.js's end-of-line flags with baseline changes, separates pages with a blank line and hands you the whole document at once as a file. It also reports the character count and page count, so you can sanity-check that nothing was skipped. For a two-page memo the difference is small. For a 300-page manual, or a batch of documents you want to search or script over, it is the difference between a minute and an afternoon.

PDF to text extracts the visible text printed on the page and nothing else. A URL printed on the page comes through as text, but a PDF link annotation attached to the words "click here" does not — the destination lives in an annotation object this extraction does not read, so you get the words and lose the target. Running headers, footers and page numbers are part of the page's text, so they appear on every page in the output; strip them afterwards if they are noise. Footnote text is extracted where it physically sits on the page, at the bottom, not inline at its reference marker. Images, form-field values and review comments also live outside the text layer and are not included.

Copy text puts the complete extraction on your clipboard, not the truncated preview. The preview box is capped at 2,000 characters purely so the page stays responsive on a long document — the full string is held in memory, and both Copy text and Download .txt use all of it. The button confirms with "Copied" and the status line says so too. Clipboard access needs a secure context, which the site provides, but a browser can still refuse if the tab lost focus at the moment you clicked; if nothing pastes, click back into the page and press it again, or use Download .txt, which never depends on clipboard permissions at all.

The extracted text exists only in the memory of the tab you are using. Your PDF is read from disk by the browser, parsed by pdf.js inside the same page, and the resulting string sits in the component's state until you press Reset, close the tab or navigate away — it is not written to local storage, not sent to an analytics endpoint and not uploaded, because the tool has no upload path. That matters more here than in most converters, since the thing being produced is the readable content of the document. The tool is free with no account and no length limit; the practical ceiling is how much text your browser is happy to hold in one string.