Extract the text from a PDF, page by page

Pull the text layer into a plain .txt file. Fast, private, and honest about the limit: if the page is a scan, there is no text to pull — that is an OCR job, not a conversion.

Text only Output: UTF-8 .txt No text layer = no output

What this reads — and what it cannot

It reads the PDF's text layer, the same stream that lets you select text in a viewer. Reading order is inferred from glyph positions, so multi-column pages and tables may come out interleaved. A scanned page has no text layer at all and will produce an empty result.

What “the text layer” actually is

A text-based PDF does not store paragraphs. It stores glyphs at coordinates, plus font references. When you select a sentence in a viewer, the viewer is reconstructing words from those positions. Extraction does the same thing, which is why the result is usually right and occasionally odd: a ligature may arrive as a single character, a hyphenated line break splits a word, and two columns may interleave line by line because the extractor sorts by vertical position across the whole page.

Knowing this saves time. If the output looks scrambled, the source is a layout the extractor cannot linearise — go to Word conversion, or select and copy from the viewer page by page.

How to extract text

Add the PDFThe preview shows the page count and a note if the file looks image-only.
Choose the format detailsKeep page breaks if you need to cite pages; add page markers if the text is going into a script or a review.
ExportYou get a UTF-8 .txt file with the text in document order.

When the output is empty

An empty file almost always means a scanned document: a picture of a page inside a PDF. There is no hidden text, so nothing can be extracted without OCR — optical character recognition — which guesses characters from pixels and needs a different engine (and a language model) entirely. Check the page in a viewer first: if you cannot drag-select a sentence, we cannot extract it either.

Item	Behaviour
Source	PDF text layer (glyph stream)
Output	UTF-8 plain text, one file
Reading order	Inferred from positions — columns may interleave
Scanned PDFs	No output; requires OCR
Formatting	None — bold, tables and images are not represented

FAQ

Can it read a scanned PDF?

No. There is no text to read. You need an OCR tool, and even then the result should be proofread — OCR confidently turns “rn” into “m” and “0” into “O”.

Why are words split across lines?

Because the PDF hyphenated them at the line break. docu-\nment is what the page actually contains; rejoining hyphens automatically would corrupt real compound words, so we leave them.

Does it keep the layout?

No — plain text has no layout. If you need columns and tables preserved, use PDF to Word and accept an approximate rebuild, or keep the PDF.