How to OCR a PDF: Pull Real Text Out of Any Scanned Document
You open a scanned contract, press Ctrl+F, search for a name you know is on page four, and get nothing. The document looks like text. To the software it is a photograph. That is the exact problem a PDF OCR tool solves: optical character recognition looks at the picture of a page and works out which letters it is looking at, so you can finally extract text from a scanned PDF instead of retyping it. This guide covers how to OCR a PDF free, which languages are supported, what you get back, and the one thing that decides whether the result is excellent or unusable.
1. What OCR Actually Does to a PDF
There are two completely different things people call a PDF, and almost every frustration in this area comes from not knowing which one you are holding.
A digital PDF — exported from Word, a design tool or a web page — carries a text layer. Under the visible page there is a real list of characters with real positions, which is why you can select a sentence, copy it, and search the file. The ISO 32000-1 specification that defines PDF treats that text layer as optional, and that single word explains everything that follows.
A scanned PDF has no such layer. A scanner, a photocopier or a phone camera produced an image of a page and wrapped it in a PDF container. Your eyes see paragraphs; the file contains pixels and nothing else. Search finds nothing, selection highlights nothing, and a converter looking for text finds an empty shelf. This is why people end up searching for why can’t I copy text from my PDF or PDF text not selectable — the text was never there.
Optical character recognition is the bridge. An OCR scanner examines the shapes on each page, matches them against learned letterforms, and outputs the characters they represent, mapped to standard Unicode code points so accented letters, Arabic script and Devanagari all come out as themselves rather than as question marks. The result is real text you can copy, search, translate, paste into an email or feed into anything else. That is the entire job of the tool at the top of this page.
2. How to OCR a PDF Online Free (Step-by-Step)
No account, no email, no trial countdown. To convert a scanned PDF to text with the tool above, four steps:
Select your scanned PDF
Drop one .pdf onto the panel above or click to browse. Non-PDF files are turned away at the dropzone rather than failing partway through a run. It is one document per run, so pick the scan you want first.
Choose the document language
Pick the language the document is actually written in from the dropdown. This matters more than people expect, and section 4 explains why.
Run the extraction
Press the button. A progress readout names the page being rendered and the page being recognised, with a percentage beside it, so a long document never looks like it has stalled. A run goes through to the last page once it starts, so if you are unsure about the scan quality, try a short document first.
Copy or download the text
The recognised text appears in a panel you can read and scroll. Press copy to put it on the clipboard — paste it once to confirm it landed — or download it as a .txt file named after your original document, which is the dependable route for anything long.
Every page of the document is scanned, not just the first, and the output is delimited page by page so you always know which paragraph came from where. There is no watermark and no daily counter: one document per run, and as many runs as you care to do, which is the usual catch with a free tier elsewhere.
3. Searchable PDF vs Extracted Text: Which One You Get
This is the most searched question in the whole category and it deserves a straight answer rather than a slogan, because the two outcomes people ask for are genuinely different things.
What this tool gives you
- The full recognised text of every page, on screen
- Page-by-page delimiters so nothing loses its place
- A copy button for the whole text in one press
- A .txt download named after your original file
- Text you can search, edit, translate or reuse anywhere
What it does not do
- Rewrite the original PDF with an invisible text layer
- Return a .docx, .xlsx or formatted output file
- Accept a JPG, PNG or photo on its own
- Take more than one document per run, or queue a batch
- Stop a scan once it has started, or resume a part-finished one
- Reproduce columns, tables or images as layout
- Open a password-protected or encrypted PDF
So if you searched for make a scanned PDF searchable and what you actually need is the words — to quote them, index them, translate them, paste them into a case file or check whether a clause exists — this tool is exactly the right one and it is free. If you specifically need the original PDF handed back with a hidden text layer welded underneath the scan, that is a different operation and we would rather tell you now than after you have waited for a scan to finish.
One related point worth knowing: if your PDF is not a scan — if you can already select the text with your cursor — you do not need OCR at all. Send it to the PDF to Word converter instead, which reads the existing text layer directly and rebuilds it as editable paragraphs. That route is faster and more accurate, because nothing has to be guessed from pixels.
4. OCR in Nine Languages
Recognition is not language-neutral. The engine is deciding between letterforms that look alike, and knowing which alphabet and which vocabulary it is reading changes what it picks. Run a French document under English and accented characters get mangled; run an Urdu document under English and you get nonsense. Choosing the right language in the dropdown is the cheapest accuracy improvement available to you.
The tool above supports nine recognition languages:
Arabic, Urdu and Hindi OCR
Right-to-left and Indic scripts are where most free tools quietly give up, so it is worth calling out that Arabic OCR, Urdu OCR and Hindi OCR are all selectable here at no cost. Connected Arabic and Urdu letterforms are harder to recognise than separated Latin type, and a clean scan matters even more for them than it does for English, but the support is real and the output comes back as proper Unicode you can paste into any editor.
One document, one language per run. A genuinely bilingual page — an English contract with Urdu annotations, say — is best handled by running it twice and keeping the good half of each result, or by pulling the relevant pages out first with the PDF splitter and scanning each part under its own language.
5. Scan Quality: The One Thing That Decides Accuracy
No optical character recognition on earth reads what is not legible. If you can barely make out a word yourself, the engine will not do better. The difference between an almost-perfect result and a page of gibberish is almost never the tool — it is the source. A few habits that change the outcome dramatically:
- Scan at 300 DPI or higher. Below roughly 200 DPI the strokes that separate an “e” from a “c” stop existing, and no amount of processing invents them back.
- Keep the page straight. A few degrees of skew costs real accuracy, and nothing downstream can undo it — re-scan the page squarely if you still have the paper, because a tilted line of type is the single most common cause of a mangled result on an otherwise clean document.
- Turn sideways pages the right way up. That is a different problem from skew, and a fixable one. If a page came out of the scanner a full 90 or 180 degrees out, the PDF rotate tool turns it upright in quarter turns and saves the corrected file, and you scan that here instead.
- Light it evenly. Phone photos are the worst case: shadow across the gutter, glare on glossy paper, one corner brighter than the rest. Flat, even light beats an expensive camera.
- Prefer printed type to handwriting. Recognition is trained on typeset letterforms. Handwritten notes, signatures and stylised display fonts are unreliable and should be treated as a bonus rather than an expectation.
- Crop away the noise. Black scanner borders, a desk edge or half the facing page all give the engine extra shapes to misread. The PDF crop tool trims them out first.
And always proofread. Even a clean scan will occasionally swap a 1 for an l or an O for a 0. For a quotation, an invoice figure or a legal reference, read the output against the page before you rely on it. Anyone advertising a guaranteed accuracy percentage on arbitrary documents is describing a marketing number, not a measured one.
6. What to Do With the Extracted Text
Plain text is the most portable thing a document can become, which is why it is the output here. Once you have it, everything downstream is easy:
- Editing. Paste it into Word, Google Docs, LibreOffice or Pages and format it however you like. This is the practical answer to scanned PDF to Word — get the words out, then style them once rather than fighting a converter’s guess at your layout.
- Searching an archive. Run your scans through one at a time, keep each .txt file beside its PDF under the same name, and a folder of dead images becomes grep-able, indexable and quotable. There is no queue here, so this is a working session rather than a batch job — but it only has to be done once per document, and the .txt files are the thing you search from then on.
- Translation and summarising. Text pastes cleanly into any translation or assistant tool. Images do not.
- Data entry. If the scan is a table of figures, the recognised text gives you the numbers to work from. For PDFs that already carry a real text layer, the PDF to Excel tool is the better starting point because it is built around grid structure.
- Accessibility and compliance. A scan is invisible to a screen reader. Extracted text is not, so recognising a document is often the first step in making it usable at all.
If the pictures matter as much as the words — figures in a paper, stamps on a form, photographs in a report — the extract images from PDF tool pulls those out separately, and the PDF to JPG converter turns whole pages into images you can drop into a slide deck.
7. Why Cortexis Is the Best Free PDF OCR Tool
Plenty of sites will OCR a PDF online. Most make it deliberately irritating until you pay: three pages free, a watermark, an email wall, a daily counter. Here is what is different, stated plainly enough that you can check every line yourself.
Genuinely free, no account
No sign-up, no email address, no credit card and no trial that expires. What you are using is the whole product.
No page limit, no file size cap
There is no page limit and no file size cap coded into the tool — though a very long document is a long run, so work in sections if you are scanning hundreds of pages. The PDF splitter breaks a big report into parts you can scan one after another.
Nine recognition languages
English, Spanish, French, German, Italian, Portuguese, Arabic, Urdu and Hindi — including the right-to-left and Indic scripts most free tools skip.
Your documents stay private
Files are never stored on our servers, never retained after your session, and never used to train or profile anything.
Progress you can see
A live readout names the page being rendered and the page being recognised, with a percentage, instead of an anonymous spinner that tells you nothing about where a long document has got to.
Clean, usable output
Every page recognised and delimited, then a copy button or a .txt download named after your file. No watermark, no stamped footer, no locked preview.
It sits inside a wider set of free document tools built on the same terms, so a scan you have just recognised can also be merged, reordered or compressed without paying for any of it.
8. Frequently Asked Questions (FAQ)
How do I OCR a PDF for free without installing anything?
Open the tool at the top of this page, drop in one scanned PDF, choose the document language and run it. The recognised text appears on screen with a copy button and a .txt download. There is no sign-up, no watermark and no cap on how many documents you run — they simply go through one at a time rather than as a batch.
Why can I not select or search the text in my PDF?
Because the page is an image, not text. Scanners, photocopiers and phone cameras produce a picture of the page wrapped in a PDF container, so there is nothing for search or selection to find. Optical character recognition reads those pixels and gives you the characters back.
Does this give me back a searchable PDF file?
No, and it is worth being precise about that. It gives you the recognised text: on screen, with a copy button, and as a .txt download. It does not rewrite your original PDF with a hidden text layer inside it. If the words are what you need, this is the right tool; if you specifically need the PDF itself rebuilt with a text layer, that is a different operation.
Can I stop a scan once it has started?
Not from the panel — once a run begins it works through to the last page, and there is no pause or resume. In practice that only matters on very long documents, which is the reason to split a hundred-page report into sections and scan them one part at a time rather than committing to the whole thing in one go. A short test document is also a cheap way to check the scan quality and the language setting before you start something big.
How long does a long document take?
It depends on the page count and how dense each page is, and we will not quote a figure we have not measured on your file. What we can tell you is the shape of it: pages are handled one after another in order, so the time grows with the number of pages rather than staying flat, and the progress readout names the page it is on so you can judge the pace after the first few. For anything in the hundreds of pages, split it first.
Which languages can it recognise?
Nine: English, Spanish, French, German, Italian, Portuguese, Arabic, Urdu and Hindi. Pick the one your document is actually written in before you run it, because recognising a French or Urdu page under English settings is the most common cause of a poor result.
Can I OCR a JPG, PNG or a photo of a page?
Not directly — the dropzone takes .pdf files and nothing else, and the only output is plain text. If your source is a photo or an image, place it into a PDF first and then run it here.
Are my scanned documents private?
Yes. Your files are never stored on our servers, never retained after your session, and never used for training or analytics. Nothing you scan is kept, queued or logged.
Working through a pile of scans? The PDF splitter breaks a long document into manageable parts you can scan one at a time, the PDF rotate tool turns pages that came in sideways back upright in quarter turns, and the PDF compressor brings a heavy scan under an email limit once you are done. If you would rather understand how machines learn to read letterforms than simply use a tool that does, our free AI and machine learning essentials course covers the pattern-recognition ideas that sit underneath optical character recognition.