PDF to Text

Two kinds of PDF, and they behave very differently

You need the words out of a PDF. Selecting them by hand works until the document is sixty pages long, or until you drag across a paragraph and the cursor slides over it like glass.

That second case is worth understanding before you start, because it decides what you get.

Some PDFs carry a real text layer. The characters are stored as characters, sitting invisibly behind the page. Pulling them back out is fast and exact: every letter comes back the way it went in.

Other PDFs are pictures. A scan, a photo of a receipt, a fax forwarded four times. There are no characters in the file at all, only pixels arranged to look like writing. Nothing can read those exactly. They have to be recognised, and recognition guesses.

This tool does both, and it always marks which one it used.

The three modes

Auto is the default. It reads the text layer of every page you asked for and reaches for OCR only on a page that comes back nearly empty, meaning fewer than eight actual characters. On an ordinary digital PDF, OCR never runs.

Text layer only never runs OCR at all. It is instant, and it returns a scanned page blank rather than inventing something to fill it.

OCR all pages forces recognition on every page. Reach for it when a PDF's own text layer comes back garbled, which some export tools produce, and you would rather trust the picture than the file.

You can run the whole document, or a range typed as 1-3, 5, 8-10.

What OCR is good at, and what it is not

Optical character recognition reads letters out of an image, so the result is a best guess rather than a transcript.

Clean, straight, printed scans come back well. Faint print, handwriting, multi-column layouts, tables, and pages scanned at an angle come back worse. Treat OCR output as a draft to proofread, not a copy to trust.

It recognises English only for now. Other scripts come back as noise.

It is also slow next to reading a text layer, a few seconds per page, and the progress line names the page it is working on so you can see it has not stalled.

What you get, and what this page downloads

The result lands in a box, split page by page with a --- Page 4 --- heading before each one. Pages that came from OCR carry an OCR badge above the box, so exact text and recognised text are never mixed up silently. Copy the whole thing to your clipboard, or download it as a .txt named after your PDF.

Your PDF is never uploaded. It is read in your browser's memory, and no PDF or extracted text is written to a server, an account, or local storage.

The page itself does fetch engines, all from noadstools.com and nowhere else. Opening a PDF pulls in pdf.js, roughly 1.73 MB. A page that actually needs OCR pulls in Tesseract, roughly 22 MB with the English data included. That is a large download to produce a text file, which is exactly why it does not happen unless a page needs it. Open a normal digital PDF and it is never requested.

There is a smaller third case. A PDF built on an unusual character encoding, or on one of the standard PDF typefaces, makes pdf.js request the matching character maps or font data, again from this origin, and only for the documents that need them.

None of that is bundled into the page, so no step runs before its engine has arrived, and even a warmed-up tab can still need font data for the next document. Opening this with the network switched off will not get you an extraction.

Watch it if you want. Open the DevTools Network tab and extract something: you will see those engine files arrive from this site, and you will not see your PDF go anywhere.

The source is public on GitHub under the AGPL-3.0 license, and the privacy page repeats those engines and their sizes in one place.

Questions

Is there a free PDF to text converter online?

Yes. This page is one.

No signup, no email, no install, no page limit, and nothing to pay. Drop in a PDF, press Extract text, and copy or download the result.

The reading happens on your device. The only thing that travels over the network is the engine code this page loads from its own site.

Is my PDF uploaded to a server?

No.

The file is opened in your browser, and both the text-layer reading and the OCR run there. No copy is kept, and nothing is stored between visits except your theme preference.

You can confirm it in your browser's DevTools Network tab while an extraction runs. The engine files come down from noadstools.com. Your PDF does not go up.

Does it work on scanned PDFs?

Yes, through OCR.

A scanned page has no text layer, so there is nothing to extract. The page is redrawn as an image and the OCR engine recognises the letters in it.

Leave the mode on Auto and scanned pages are handled without you doing anything. Choose OCR all pages if you want recognition run on every page regardless.

How accurate is the OCR?

Approximate. That is the honest answer for any OCR, not a limitation of this one.

A crisp, straight, printed scan usually comes back close to perfect. Faint ink, handwriting, dense tables, and skew all cost accuracy, and a badly skewed page can come back mostly wrong.

English only for now. Proofread anything a page badge marks as OCR.

Why is the OCR download 22 MB?

Because recognition needs a trained model, and the model is most of that size. The English language data is bundled into it.

It comes from noadstools.com, not from a third party, and it downloads once and only when a page genuinely needs OCR. A PDF with a working text layer never triggers it.

If you are on a metered connection and want to be certain it stays untouched, pick Text layer only before extracting.

A page came back empty. What went wrong?

Probably nothing. In Text layer only mode, a scanned page has nothing to give, so it comes back blank on purpose. Switch to Auto or OCR all pages and try again.

If OCR ran and still returned nothing, the page image may be too faint or too complex to recognise, or the page may genuinely be blank.

The note under the output says which of those happened rather than leaving you to guess.

Can it open a password-protected PDF?

No. An encrypted PDF cannot be read without its password, and this tool does not ask you for one.

Remove the protection in whatever application you normally open it with, then add the unlocked file here. The tool tells you the file is locked instead of pretending it is corrupt.

Related tools

PDF to JPG | turn PDF pages into JPG or PNG images. Split PDF | pull out the pages you want before extracting. Merge PDF | combine PDF files in any order. Sign PDF | draw or type a signature and place it on a page. Watermark PDF | stamp text or a logo across the pages.