The question underneath this one
Most people arrive at a tool like this having already tried to select the text in a PDF and got nothing at all, or got a jumble, and concluded that something is broken.
Nothing is broken. There are two completely different things both called a PDF, and only one of them contains any words.
A document with a text layer
This came out of Word, a drawing package, an accounts system or a website. It stores the actual characters, each with a position on the page. Everything in it can be selected, searched, copied and read aloud by a screen reader, and pulling the words out is instant and perfectly accurate, because the words are simply there.
A scan
This is a photograph of paper. Somebody put a sheet on a scanner or took a picture of it with a phone. Your eye reads the words on it, but the file contains no letters whatsoever, only millions of coloured dots arranged in the shape of letters.
Selecting text on a scan gives you nothing because there is nothing to select. No tool can copy words out of it, because there are no words in it, and that is true of the expensive ones too.
This tool checks which of the two you have before it does anything, and tells you. If your PDF is a scan you get a straight answer and the reason, rather than an empty box and an afternoon wondering what you did wrong.
How this tool works
A PDF does not contain lines or paragraphs. It contains pieces of text, each with a position, in whatever order the program that made it happened to write them, and frequently with no spaces between them: the word "Balustrade" can arrive as "Bal", "ust" and "rade".
So the pieces are grouped by height into lines, sorted left to right within each line, and a space is inserted only where the horizontal gap is wide enough to be a real one, measured against the type size rather than a fixed distance. That single step is the difference between a useful extractor and the wall of jumbled words most of them produce.
What happens to the line breaks
Every other extractor gets this part wrong. It is worth understanding, because it decides whether the output is useful or a mess.
A PDF does not contain paragraphs, sentences or even lines. It contains pieces of text, each with a position, written in whatever order the program that made it happened to use. Very often there are no spaces between them either: the word "balustrade" can be stored as three separate pieces reading "bal", "ust" and "rade".
So the pieces have to be put back together by working out which ones share a line, sorting them left to right, and deciding where the gaps between them are wide enough to be real spaces rather than the normal spacing between letters. Get that last judgement wrong in one direction and you get "bal ust rade". Get it wrong in the other and two columns of a table run into one another.
Then there is a second decision, which is yours rather than the tool's:
- Keep the layout gives you one line out for every line on the page. This is what you want for a table, a schedule, an address block, a parts list or anything where the arrangement on the page is carrying meaning.
- Join into paragraphs puts lines back together where they were only broken by the page running out of width. This is what you want for prose you are going to paste into an email or a document and let it wrap on its own. Without it, text pasted from a PDF breaks in all the wrong places, the dead giveaway of copied and pasted PDF text.
What you can do with the result
Copy the whole thing to the clipboard in one press, or save it as a plain .txt file. Plain text opens in anything, on any machine, forever, and carries no formatting, no fonts and no metadata with it.
There is also an option to mark where each page starts, which is worth turning on for anything you might later need to quote a page number from. Once the text is out of the document, the page boundaries are the first thing lost.
The documents people want the words out of
Every free extractor that works by upload has your document on somebody else's machine, and the documents people want the words out of are contracts, tenders, reports, letters from a solicitor and statements from a bank.
This one reads the file on your own device. There is no account, no size limit, no watermark and no request leaving your browser. Open your network tab and watch while it works.
Common questions
Why did copying and pasting from my PDF give me nothing?
Because there is almost certainly no text in it. A scanned document is a photograph of paper: your eye reads words, but the file contains coloured dots and nothing else. There is nothing to select and nothing to copy. Drop the file in above and it will tell you within a second which kind you have.
Does this recognise the text in a scan?
Not this tool, but there is one here that does: Make a scan searchable, which runs character recognition entirely in your browser. This tool reads text that is already in the file, which is why it is instant and exactly right rather than a good guess. Recognition is the other way round: it works on a picture and produces its best reading, so it can help where this cannot, and it should be checked where the wording matters.
Why is my text coming out with the lines in odd places?
Try "Join into paragraphs". A PDF has no paragraphs in it, only lines at positions, so line breaks in the output are wherever the original page happened to run out of width. Joining them puts the text back into paragraphs that will wrap properly wherever you paste them. Keep the layout is the right choice for tables and schedules, where the arrangement is the information.
Will the table I extracted still line up?
The rows will be on their own lines and the columns will be in the right order, which is usually enough to work with. It is text, though, so the columns are separated by spaces rather than held in a grid. If you need it back as a real table you will have to paste it into a spreadsheet and split it.
Is my PDF uploaded anywhere?
No. It is read on your own device and the text never leaves it. That matters more here than on most tools, because the sort of document somebody wants the words out of is usually a contract, a report or a letter.
Is there a page limit?
None. A long document takes a moment because every page is read one at a time on your own machine, and there is a page counter while it works. Nothing is capped, watermarked or held back for a paid version.