PDF to Text
Pull the text out of a PDF and get clean, readable output. Lines that the page layout chopped up are merged back into paragraphs, missing spaces are restored, and words split across a line break are joined. Everything runs locally — your file never leaves your device.
Drop a PDF here
or choose a file — nothing is uploaded
One PDF, up to 100 MB. Scans without a text layer cannot be extracted.
Extraction runs entirely in your browser — your PDF never leaves your device. Text is read in the order it appears on the page. Tables and multi-column layouts may come out scrambled. Vertical Japanese text reads top to bottom, right to left, so the extracted order may come out reversed.
How to use
- Drop a PDF into the box or click “Choose PDF”. The file stays on your device.
- Pick an output format: plain text, Markdown with page headings, or structured JSON.
- Open the options panel if you need it — turn hard-wrap removal, space cleanup or repeated header/footer removal on or off.
- Watch the per-page progress finish, then copy the result or download it as .txt, .md or .json.
- Only need a few pages? Enter a range such as 1-5, 8 in the page range field.
Details
Copy text straight out of a PDF and you usually get something unusable: one sentence sliced across five or six lines, words fused together, and a stray blank line in the middle of a paragraph. The reason is that a PDF does not store text the way a word processor does. It holds a stream of glyphs, each one positioned by a coordinate — there is no notion of a word, a space or a paragraph to copy. Every tool that produces readable text has to reconstruct those ideas from the geometry, and how well it does so is the whole difference between a garbled blob and a clean document.
This tool rebuilds the layout explicitly. Glyphs on the same baseline are clustered into lines by their vertical coordinate, and the horizontal gap between neighbours is compared against the average character width to decide where a space belongs — that is what prevents the classic “textbecomesonewords”. Lines broken only by the page width are merged back into paragraphs, and a word split across a line break is rejoined: inter- plus national becomes international. A hyphen that is part of the content is kept instead, so COVID- plus 19 stays COVID-19 and Anglo- plus Saxon stays Anglo-Saxon. With repeated header/footer removal switched on, lines that recur on most pages are dropped automatically. If the extracted text still needs tidying, send it to the text sorter to sort or deduplicate lines, or to the word counter to count words.
A raw PDF is a poor input for an AI model, and the shape of the output matters. Plain text is the simplest thing to paste into any chat or prompt. Markdown keeps a “## Page N” heading before each page, so the model — and you — can still tell where a page break fell. JSON returns the same content as structured data, with the page number and text of every page aligned, which is what you want when a script processes the result. All of it happens locally, on your own machine.
FAQ
Can it read a scanned PDF?
No. A scanned PDF is a set of images with no text layer, and this tool extracts the text layer only — there is nothing to extract. It detects that situation and tells you so, then points you to <a href="/pdf-to-images/">PDF to Images</a> so you can export the pages and run OCR elsewhere. OCR is not part of this tool.
Will tables come out as tables?
Not in this first version. Text is extracted in reading order, line by line, so a table’s cells come out as text separated by spaces and line breaks rather than as a Markdown table. The cell contents are preserved; the table structure is not.
Is my PDF uploaded anywhere?
No. The file is opened and parsed entirely in your browser. Nothing is sent to a server, nothing is stored, and the tool never makes a network request with your document. That is deliberate: contracts, reports and invoices are exactly the documents people want to feed to an AI, and they often contain personal data.
Does it support Chinese, Japanese and Korean PDFs?
Yes, as long as the PDF carries the needed character maps. The tool ships the CJK cmaps and standard fonts that many extractors omit, which is why some tools return empty text for these documents. Horizontal Japanese works normally; vertical text may come out in the wrong order, and the interface says so.