PDF to Text: Extract Text From a PDF With the Columns in the Right Order
Pull the text out of a PDF with two column pages read in the right order, repeated headers dropped, and split words rejoined. Nothing is uploaded.
- Free, no account
- No watermark
- No usage limit
About the PDF to Text
You download a paper, select the whole thing, paste it into your notes, and what lands is nonsense. First line of the left column, first line of the right column, back to the left again. Every sentence has been chopped in half and welded to a sentence from somewhere else on the page, and cleaning that up by hand takes longer than retyping the document from scratch.
Your reader is not misbehaving, and the file is not broken. A PDF does not hold text the way a document does. It holds fragments of characters and the coordinates they were painted at, and nothing in the file records which fragment comes after which. So every converter has to guess the reading order, and nearly all the free ones guess the same lazy way, by sorting on how far down the page a fragment sits and reading straight across. That works fine on a single column page. Point it at a journal article, a newsletter or a court filing and you get the mess above.
PDF to Text works out the layout before it reads a word. When a strip of white space runs down the page and the text respects it, that is a gutter, and the page gets read one column at a time rather than straight across. A headline or a footer that crosses the full width stays where it belongs instead of being dragged into the top of column one. The counter on screen tells you what it found, so 7 pages set in columns explains why those seven read differently.
Two more things happen while it reads. The running head and the page number that repeat on every sheet get dropped, since their only real talent is interrupting a sentence halfway down your text file. And a word broken across a line ending gets put back together, so informa- and tion arrive as one word instead of two.
Nothing is uploaded. The file is read on your own computer, which counts for a lot when the document is a settlement agreement or a scan of your own medical notes. No sign up, no allowance of two files a day before it wants you on a plan, no watermark on the way out.
How to get the text out of a PDF
- Drop the file in. Drag it onto the box, click to browse, or paste it straight from your clipboard. It starts reading immediately and counts the pages off as it goes.
- Read the notes under the counters. They say how many pages were set in columns, how many repeated header lines were dropped, and whether any of the pages hold no selectable text at all.
- Pick the shape that suits what comes next. Clean prose puts the paragraphs back together for pasting into a document or an AI chat. Keep line breaks leaves them exactly where the PDF had them, which is what you want for verse, code, or anything you plan to compare line by line. Keep layout rebuilds the page on a character grid so a table still lines up.
- Narrow it down if you only want part of it. The pages box takes
1-3, 8, 12-and only reads the pages you name, which is how you reach the back half of a long book. Switching to page by page gives every page its own box with its own copy button. - Copy the lot, or download the .txt. Both take exactly what is on screen, with your options applied.
Why the paragraphs need putting back together
A PDF has no paragraphs in it either. What it has is lines, and a line ends where the typesetter decided it should end. Copy a page straight out of a reader and you get a hard break after every one of them, which is why pasted PDF text arrives ragged and refuses to reflow when you change the margins.
Clean prose mode works the breaks out from the geometry instead. A bigger than usual vertical gap starts a new paragraph. So does an indent at the start of a line, a bullet, or a previous line that stopped short and ended on a full stop. Everything else gets joined back into running text. That is a judgement call rather than a lookup, so a heavily designed page will occasionally break somewhere odd, it is the trade for text you can actually paste.
The column reading has a limit worth knowing about too. It only fires when the gutter is a real column gutter, wide and respected down the whole height of the page. A label and value form, or a numbered list where the numbers sit in their own narrow strip, is left alone on purpose, because reading those one side at a time would separate every label from its answer.
The one thing this cannot do
If the PDF is a scan, there is no text in it to extract. The page holds a picture of words, and no amount of clever reading order changes that. You get a plain message saying so rather than an empty box and no explanation. Pulling text off a scan needs character recognition, which is a different job and not one this tool does.
Frequently asked questions
Some of the text came out as gibberish symbols, why?
That one is the file's fault rather than the reader's. To keep the size down, a PDF can embed a cut down copy of a font and skip the table that says which letter each shape stands for. Your reader still draws the page perfectly, because it is drawing shapes, but there is genuinely no letter recorded anywhere to copy. Nothing can extract that text, and a page image plus character recognition is the only way round it.
Can I get text out of a PDF that will not let me copy?
Usually yes. A PDF that blocks copying carries a permissions flag, and that flag is a request to the reader program rather than a lock on the content, so the text is sitting there in the open for this to read. A PDF that demands a password before it will open is a different animal, the content really is encrypted, and you will get an error back instead.
Is this the same as PDF to Word?
No, and the difference matters if you were hoping to keep the formatting. A Word conversion tries to rebuild headings, tables, images and fonts as editable objects, and it usually gets a slice of that wrong in ways that take you a while to notice. This gives you the words and nothing else. If the next step is a search, a summary, a translation or a prompt, plain text is the better input anyway.
What happens to tables?
Clean prose flattens them, which is honest but not much use. Keep layout is the mode for tables. It puts every cell back at roughly the column it sat in using spaces, so the shape survives a paste into a plain text editor. Getting a genuine spreadsheet out of a PDF table is a harder problem and this makes no claim to solve it.
How big a file can I put through?
Five hundred pages in one pass, and that ceiling is memory on your own machine rather than a plan you have not bought yet. Text heavy documents move quickly, a few seconds for a hundred pages is normal. A longer book still goes through, it just arrives in stretches. Put 501- in the pages box and it reads that stretch instead, and the note on screen tells you where you are.