Make a scanned PDF searchable with OCR
Run OCR on a scanned PDF for free in your browser, with no upload. The recognized words are added as an invisible layer over the scan, so the page looks as it does now and Find starts working; a plain text file is offered too.
Read the words off each scanned page and add them back as an invisible layer, so the scan looks the same and Ctrl+F starts working.
In detail
OCR PDF reads the words off a scanned page and writes them back into the document as an invisible text layer sitting on top of the ink. The page itself is copied through byte for byte, so the scan looks identical and nothing is re-compressed; what changes is that Find works, text can be selected, and the file stops being a picture as far as software is concerned.
Everything happens in this tab. Tesseract, compiled to WebAssembly, is downloaded from File4Edit the first time and kept by your browser afterwards, and each page is drawn by the same PDF renderer the rest of the site uses. Recognition is a guess, so a plain .txt of the words is offered beside the PDF and the wording on the page says where that guess is weak.
Make a scanned PDF searchable with OCR at a glance
- Input
- One scanned PDF
- Output
- The same PDF with an invisible text layer, and a .txt of the words
- What it does
- Invisible text layer over the untouched scan
- Plain .txt of the recognized words
- Six languages, each sized before you fetch it
- 200, 300 or 400 DPI
- Every page or a range you type
- Handles pages the scanner saved rotated
- Warns about pages that already carry text
- Who it is for
- Anyone with a scan, a fax or a photographed document that Find cannot search.
- Limits
- Recognition is a guess: handwriting, angled photos and faxes come out poorly.
- Six languages are installed, and each is a separate download of up to 2.82 MB.
- Sixty pages per run, and a long run keeps the tab busy for minutes.
- A page that already has text should be left out, or it ends up with two layers.
- Privacy
- Runs entirely in your browser: the PDF is never uploaded, and nothing is kept after you close the tab. The engine and language files are served by File4Edit, not by a third party.
How to make a scanned PDF searchable
- Open the scan Drop the PDF onto the page. Its page count and page size appear, along with a rough estimate of how long the run will take and which pages already carry text.
- Pick the language English, Spanish, Italian, Portuguese, German or French. Each is a separate download and the size of each is shown in the list before anything is fetched.
- Set the detail and the pages Choose 200, 300 or 400 DPI, and either every page or a range such as 1-3, 7, 9-. More detail finds smaller type and takes proportionally longer.
- Choose what to save The searchable PDF, a plain .txt of the recognized words, or both. The filename field names them, and the text file takes the same name with .txt.
- Read the pages Press Read the pages. The engine downloads once, then each page is drawn and read and the words appear as they are found. Stop ends the run after the page it is on.
Make a scanned PDF searchable with OCR: common questions
Does adding OCR change how the scanned page looks?
No. The original page is copied through untouched and the recognized words are drawn on top as invisible text, so the scan looks exactly as it does now. What changes is that Ctrl+F finds words in it and selecting text copies them out.
How accurate is the recognition?
Good enough to search on a clean 300 DPI scan of printed text, and unreliable on much else. Handwriting, a photo of a page taken at an angle, a fax, and small or decorative type all produce mistakes. That is why a plain .txt is offered alongside the PDF: read it before you rely on the result.
Is my PDF uploaded to a server to be read?
No. The recognition engine is WebAssembly and runs inside this tab, and every page is drawn and read on your own machine. The engine and the language file are downloaded from File4Edit once and then kept by your browser, and the document itself never leaves it.
Why is the first run slow?
Two separate reasons. The engine is about 1.4 MB and a language file is between 691 KB and 2.82 MB, and that download happens once — the size is on screen before anything is fetched. The work itself is the rest: reading one 300 DPI page takes a few seconds on a single thread, so a long document takes minutes and the tab is busy throughout.
What if the PDF already has text in it?
Leave those pages out. A page that already carries a text layer would end up with two, and searching gets worse rather than better. The tool reads each page when you open the document and names the pages that already have text so you can exclude them. If a document is already searchable and you only want the words, use PDF to text.