Make a scanned PDF searchable in your browser. Your tax, medical and legal documents are never uploaded. No account, no page limit, no watermark.
🔒 Runs in your browser
The documents worth making searchable are the ones you least want to upload
A scanned PDF is a picture of a document. The words are pixels, so Ctrl+F finds nothing, and neither does your operating system's search.
The files where that matters most are rarely trivial: tax returns and notices, payslips, bank statements, medical letters, court exhibits, insurance claims. Making those searchable means handing them to whatever OCR service you found — which is the moment most people stop and think.
Here the recognition engine downloads to your browser and the work happens on your own machine. Nothing is transmitted. There is no account, no page cap and no daily limit, because there is no server doing the work and therefore nothing to ration.
Your scan is not altered
This is the part that separates a good OCR result from a destructive one.
The page images are not re-rendered, re-compressed or replaced. The recognised words are added as an invisible text layer positioned over the original scan — each word sitting exactly where its ink is, painting nothing.
So the file looks identical to the one you started with, at the same resolution and the same quality, and Ctrl+F now works. If the document might ever have to be produced as evidence, the original scan is the thing that matters, and it comes through untouched.
⚠️ Recognition is a guess, and you should treat it as one
The important consequence: never use a search of the result to prove a word is absent. A word the recogniser misread is a word your search will not find, and concluding "it isn't in there" from a failed search is how people miss things that were in there all along.
Use it to find what you are looking for. Do not use it to prove something is not there.
What this does not do
It does not correct the scan. Straightening, de-skewing and contrast repair are separate jobs; a badly scanned page recognises badly and the honest fix is to rescan it.
It does not translate or reformat. The text layer carries the words that are on the page, positioned where they are on the page.
It does not handle non-Latin script. Words it cannot encode are left out and counted rather than written wrongly.
It does not make an already-searchable PDF better. If the file has real text, this tool tells you so and stops.
This tool runs entirely inside your browser. Your file is never uploaded — it never leaves your device, so there is nothing for us or anyone else to read, store or leak.
Common questions
Is my document uploaded for OCR?
No. The recognition engine is downloaded to your browser and the whole job runs on your own machine. That is unusual - almost every OCR service uploads the file to a server, which is exactly what people hesitate over with a tax return, a payslip or a medical letter. There is no account, no queue and no page limit here, because there is no server doing the work and therefore nothing to ration.
Does this change how my scan looks?
No, and that is deliberate. The page images are not re-rendered, re-compressed or replaced. The recognised words are added as an invisible text layer sitting over the original scan, so the document looks byte-for-byte the same on screen while Ctrl+F starts working. For anything you might have to produce as evidence, the original scan is the thing that matters and it survives intact.
How accurate is it?
On a clean printed scan at a reasonable resolution, very good. On a faint fax, a skewed phone photo, a poor photocopy or anything handwritten, noticeably worse. Recognition is a statistical guess, not a reading, so treat the text layer as a search aid rather than a transcript. In particular, never use a search of the result to prove that a word is absent from a document - a word the recogniser misread is a word your search will not find.
What happens to words it cannot handle?
They are left out and counted, and the number is shown to you. Standard PDF fonts cannot encode characters outside the Latin range, so words in Cyrillic, Greek, Arabic or CJK script cannot be written into the layer at all. Leaving them out is the honest option: a missing word is simply not found, whereas a word written incorrectly is found as something it is not, which is worse in a document you are searching for a reason.
My PDF already has text. Should I run this anyway?
No, and the tool checks before you spend the download. If the pages already carry real text then the file is already searchable and recognition would only layer a second, less accurate copy of the same words underneath the correct ones. You would be making the file larger and the search results worse.