PDF to Text

Getting the words out

A PDF that was created from a document — exported from a word processor, generated by a system, saved from a browser — contains real text. This pulls it out, page by page, with a marker between pages.

The scanned-document problem

A PDF made by scanning paper contains no text at all. Each page is a photograph, and the words are pixels arranged to look like letters. There is nothing to extract, and this page will tell you so rather than handing back an empty file with no explanation.

Getting text out of a scan requires optical character recognition, which analyses the shapes of letters and guesses at them. That is a substantially harder job than reading text that is already there, and it is not what this tool does.

A quick way to tell which kind you have: open the PDF in any reader and try to select a word. If the cursor highlights it, there is text. If you get a selection box over the whole page, it is a scan.

Layout does not survive

PDF stores text as fragments placed at coordinates. There is no notion of paragraphs, columns or tables in the file — those are visual arrangements, not structure. So extraction reconstructs line breaks from vertical positions, and does a reasonable job on ordinary prose.

Where it struggles: multi-column layouts, which tend to interleave; tables, which lose their columns; headers and footers, which appear mid-flow. That is a limitation of the format rather than of the extraction, and every tool faces it.

Common uses

• Quoting from a document without retyping it.
• Searching a long PDF for something, in a text editor.
• Getting content into another application that will not accept a PDF.
• Counting the words in a document.
• Checking whether a PDF is a scan before sending it somewhere that needs real text.

Questions people ask

Why did I get almost no text?

Because the PDF is probably a scan. The pages are images of paper, so there is no text in the file to extract.

Can you read scanned documents?

No. That needs optical character recognition, which analyses letter shapes and guesses at them — a much heavier job than reading text that is already present.

Why is the layout wrong?

PDF stores text fragments at coordinates with no notion of paragraphs or columns. Line breaks are reconstructed from position, which works well for prose and poorly for tables and multi-column layouts.

How can I tell whether my PDF has real text?

Try selecting a word in any PDF reader. If it highlights, there is text. If you get a box over the whole page, it is a scan.

Is my PDF uploaded anywhere?

No. It is read and rewritten in your browser. This site has no server that could receive it.

Does it work with a password-protected PDF?

Not reliably. An encrypted file has to be unlocked before it can be changed. Open it in a reader, enter the password, and save an unprotected copy first.

Is there a file size limit?

Only your device's memory. There is no server limit and no queue, though very large files will take a moment and use a fair amount of RAM.

Your file is not uploaded

This is the whole point of the page, so it is worth being exact about it.

Every well-known site offering these tools works by sending your PDF to their servers, processing it there, and sending a result back. For a holiday itinerary that is merely unnecessary. For a contract, a payslip, a medical letter, a bank statement or a scan of your passport, it means handing a private document to a company you have never dealt with, to be held for an unspecified time on infrastructure you cannot inspect.

Nothing here is uploaded. Your browser reads the file from your disk, rewrites it in memory, and hands you the result. The site is a set of static files with no server capable of receiving a document.

The libraries doing the work are served from this site rather than borrowed from someone else's CDN, so using these pages does not make a request to any third party either. You can check that on the about page, which lists every request this site makes, read live from your own browser.

It also means it works offline once loaded, there is no file size limit beyond your own memory, and there is no queue.