How to use Searchable PDF (OCR)
- 1Drop the scanned PDF. If it is password-protected, you are asked for the password before anything is read.
- 2Pick the Language of the text. Eleven options are available, including Korean + English for mixed documents.
- 3Leave Which pages on Only pages without text (recommended), or choose Every page to read all pages again.
- 4Press Make it searchable, then download the searchable PDF, download the text as a .txt file, or press Show recognized text to check it on the page.
When you need a searchable PDF
Scanned paperwork piles up: signed contracts, lease agreements, receipts for an expense claim, old letters, a book chapter copied for study, faxes saved as PDF. They all look fine but behave like photos. Ctrl+F finds nothing, you cannot copy a clause into an email, and your computer's file search does not know what is inside. Making the file searchable fixes all three at once, without retyping and without changing how the pages look. It is also a first step before quoting from a document, translating it, or reading it with a screen reader.
What a searchable PDF is
A scanner saves each page as a picture, so a PDF reader has nothing to search or copy. This tool reads every scanned page with OCR, optical character recognition, and writes the recognized words back into the PDF as an invisible text layer placed over the picture. The page still shows the original scan, but search finds words, text can be selected and copied, and search tools can index the file.
Everything runs in your browser
The OCR engine and the language data are loaded once and then run on your device. The PDF itself never leaves your computer or phone, which makes the tool suitable for contracts, medical records and ID documents. The first run in a language takes a few seconds longer while its language file loads; after that the browser keeps it.
Pages that already have text
Mixed files are common: a few scanned pages inside a normal PDF, or a signed scan appended to a typed contract. By default the tool only reads pages that have no usable text and leaves the rest untouched, so an existing text layer is never replaced by a worse one. Choose Every page only if a scanner added a poor text layer and you want the pages read again. Be aware that this adds a second layer on top of any real text, so searches on such pages can find each phrase twice.
What happened when we tested it
We built a 2.26 MB, four-page lease: three pages scanned at 200 DPI, 1654 by 2339 pixels, slightly tilted and speckled like a real office scan, plus one ordinary page with real text. With the default setting the tool reported 3 pages read, 1 page that already had text and 1,327 words found, in about 28 seconds in headless Chromium, including loading the English engine for the first time. The searchable PDF was 2.19 MB as shown on the download button, only about 38 KB larger than the original, and the pages looked exactly as before.
About 96 percent of the words in the lease text were recognized correctly. The phrase security deposit, which appears 15 times, was found all 15 times in the recognized text. The errors were stray symbols where speckles sat in the margins, which is typical of OCR on a dirty scan. Running the same file with Every page took about 32 seconds, read all four pages and found 1,657 words, and the page that already had text now returned each line twice when searched. That is why the default skips such pages.
Accuracy and limits
Pages are rendered at up to 300 DPI before recognition. Clear scans of printed text come out close to perfect; faded copies, skewed photos, small type and dense tables lose accuracy, and handwriting is not supported. Words the engine is very unsure about are left out of the hidden layer rather than guessed, so a search miss is more likely than a wrong hit. Pick the right language, because the wrong one gives garbled words. If a page reports no readable text, it may be blank, a picture without writing or a very faint scan.
Getting better results
Scan at 300 DPI in greyscale when you can; colour adds size without helping recognition, and pure black and white can break thin letters. Straighten crooked pages with Rotate PDF first, because sideways or upside-down pages are usually read poorly. For phone photos, shoot from directly above in even light. Splitting a very long document into parts with Split PDF keeps each run short.
Frequently asked questions
Will the PDF look different after OCR?
No. The page images are kept exactly as they are. The only change is a hidden text layer that readers use for search and copy.
How much bigger does the file get?
Only a little: the hidden text plus one font, about 38 KB for three pages in our test. If you need a smaller file, run it through Compress PDF afterwards; the text layer survives compression.
Which languages can it read?
English, Korean, Korean + English, Spanish, French, German, Portuguese, Italian, Dutch, Polish and Turkish. Pick the language the document is written in.
Is my scanned document uploaded?
No. OCR runs inside your browser. Nothing from the file is sent to a server.
How long does it take?
Roughly a few seconds to ten seconds per page on a laptop, depending on the page and the device, plus a short wait the first time a language loads. Phones are slower. Stop cancels the run.
I only want the text, not a PDF
Use the Download text button here, or the PDF to Text tool, which also handles scans.

