Extract Text from PDF

A PDF does not store paragraphs, sentences or even lines. It stores instructions to place a run of glyphs at a particular point on a page, which is why naive extraction produces text that runs together or arrives in a jumbled order. Rebuilding readable text means inferring the structure back from geometry: fragments sharing a baseline belong to the same line, and an unusually large vertical gap between lines is almost always a paragraph break rather than ordinary leading. That inference is reliable for the single-column prose that makes up most documents, and it is exactly where it gets harder — two-column academic papers, tables, text wrapped around figures — that you should read the output critically before trusting it. The other thing worth knowing before you start: if a PDF was produced by a scanner or a phone camera, there is no text layer at all. The page is a photograph, the words are just shapes, and no amount of extraction will find them; that job needs optical character recognition, which is a different technology entirely.

Each text fragment in a PDF carries a position. Fragments whose vertical positions match within a small tolerance are grouped into one line and ordered left to right; a blank line is inserted where the drop between consecutive lines is more than about 1.6 times the document’s typical line spacing, which is where a paragraph break sits.

This reads the PDF’s existing text layer — it does not perform optical character recognition, so a scanned or photographed document returns nothing and is reported as such. Multi-column layouts, tables and text inside figures come out in reading order rather than visual order, and headers, footers and page numbers appear inline. Documents over 300 pages are refused. Formatting such as bold, italic and heading levels is not preserved.

A 12-page article extracts to roughly 6,000 words of plain text with its paragraphs intact, ready to paste into a document or a word counter.

1. Choose your PDF file.
2. Decide whether to keep line breaks or produce continuous text.
3. Extract, then read the preview.
4. Copy the text or download it as a .txt file.

• Expecting text from a scanned document, which contains images of words rather than words.
• Trusting the reading order of a two-column paper without checking it.
• Assuming tables keep their rows and columns — they arrive as sequential text.
• Forgetting that repeated headers and footers appear once per page in the output.

Pull the words out of a PDF and keep them. The text layer is read page by page, line breaks and paragraphs are reconstructed from where the words actually sit on the page, and the result can be copied or downloaded as a plain .txt file.

Students quoting sources, researchers processing papers, and anyone whose PDF reader refuses to let them select text properly.

• Paragraph structure is rebuilt from the page layout, not dumped as one run-on block.
• Word and character counts come with the extraction.
• Tells you plainly when a PDF is a scan and therefore has no text to extract.
• The document never leaves your device.

• Quoting from a paper or report without retyping it.
• Getting text out of a PDF whose reader blocks copying.
• Feeding a document into a word counter, translator or spell checker.
• Archiving the contents of a PDF as searchable plain text.