PDF to Text

How to use

1. Drop your PDF into the Document box. Every page is read straight away, and the strip of pages shows which ones gave text.
2. Tick Add page markers to rule the pages off, or Join words split at line ends to close up words broken by a hyphen.
3. Press Copy text, or Download for a .txt named after the file. Pages marked no text are the OCR tool's job.

About this tool

Extracting text from a PDF ought to be the easy part, and with a digital PDF it is. The characters are already sitting in the file. Reading them out is a matter of walking each page and stitching the fragments back together, line by line, in the order the document stores them.

Drop a file and the reading starts at once; there is nothing to press. The words fill a read-only well as each page is read, a long document can be stopped part way, and Copy text waits until the whole of it is in. The figure counts the words and characters you are about to take, and a .txt named after the original sits beside it: contract.pdf comes back as contract.txt.

The output gets tidied on the way through. Runs of spaces collapse to one, spaces left dangling at the end of a line are dropped, and any stretch of three or more blank lines shrinks to a single gap. Pages are separated by a blank line. Add page markers heads each page with a --- Page 1 --- line instead, which is how you keep track of where a quote came from in a 200-page report.

Characters never get invented here. A scanned page is a picture rather than a store of letters, so it gives nothing; the page strip names every such page, the figure counts them, and the OCR tool is one press away. A page whose characters come back as symbols rather than words, the mark of a font with no Unicode map, is flagged the same way.

Layout is the other deliberate casualty. Columns and ruled tables flatten into one run of plain text, and so does every font choice and indent. If you want those rebuilt into a document you can keep editing, PDF to Word does that instead. If you only want the words, this is quicker and the result is easier to clean up.

Thirty-eight pages of report, quoted with page numbers

A committee report needs three passages quoted, each with a page citation. Drop it in and the text fills the well while the run counts its way to 38; tick Add page markers and the same text comes back ruled off by --- Page 1 --- lines, without reading the file again. Find each passage with your browser's own search; the marker above it is the citation.

Pages 30 through 38 are scanned exhibits, and the page strip says so: each of them reads no text, the figure counts nine pages unread, and the note under the file names them for the OCR tool. The rest is copied with one press, or saved as report.txt, named after the file it came from.

Good to know

• No text layer means no output. Scans and photographed pages come back empty, and the strip names each one so nothing is silently missing.
• Fragments are stitched in the order the PDF stores them, which is usually reading order and occasionally is not. Side-by-side columns and floating text boxes can interleave, so give a complicated page a quick read before you trust it.
• Joining split words has no dictionary: a real compound that happened to break at a line end, like well-known, joins into wellknown. That is why it is off until you ask for it.