Scanned PDF Converter: Make a Scanned PDF Searchable, and See Which Pages to Check

You are three weeks into a disagreement about a delivery date and you need the one clause that covers it. The lease came back from the solicitor as a scan, it runs to sixty pages, and Ctrl+F finds nothing at all. The clause is right there on the screen. But the file holds sixty photographs of paper, and a photograph of the word "delivery" is not the word "delivery" as far as any search is concerned.

Character recognition closes that gap, and nearly every free tool offering it wants your sixty page lease uploaded to its server first. What actually arrives as a scan is contracts, medical letters, tax returns, a copy of a passport, which is close to a definition of the paperwork you would not email to a stranger.

Scanned PDF Converter reads the pages on your own computer instead. No upload, no sign up, no allowance of two files a day before it starts asking for card details. What you get back is your original PDF with the words tucked underneath the page pictures, so a search now lands on the right line.

Recognition is a guess, and the quality swings wildly between a clean laser printed page and a third generation photocopy that went through a fax machine in 2004. Most tools hand you the searchable file and say nothing about which end of that you landed on, so you find out weeks later when a search comes back empty and you cannot tell whether the phrase is missing or was simply misread.

Every page here comes back with its own score out of 100, weighted by how much text each word carries. A long confident paragraph therefore counts for more than three specks of dirt in the margin, which is much closer to what your eyes would tell you than a plain average is. Pages land in one of three bands, and the count of pages worth checking sits at the top.

Click any page and the scan opens with a box drawn round every word the reader found, colored by how sure it was, plus a list of the fourteen it struggled with most. Proofreading a forty page scan stops being an afternoon. You look at the two amber pages, check the handful of words it flagged, and get on with your day.

Speckles on the glass, staple holes, the gray shadow down the gutter of a bound book. The reader tries to read all of it, and back comes a scattering of things like "rn" and "l1" sitting at twenty percent confidence. They are why a searchable scan sometimes reports a match in the middle of an empty margin.

So there is a slider. Anything the reader was less sure of than the number you set is kept out of the finished file. The counter beside it moves as you drag, so you can see what you are giving up, and pulling it to zero keeps every word.

1. Put the file in. Drag it onto the box, click to browse, or paste it from your clipboard.
2. Pick the language on the page. Seven are offered, and the pack for the one you choose downloads once.
3. Choose the detail. Balanced suits almost everything. Most accurate is worth it for small type or a scan that already looks rough.
4. Press read. Pages appear with their scores as they finish, so you can review the first ones while the rest are still going. Stop whenever you have seen enough.
5. Check the amber pages, set the slider, download. The plain text download is there if the words are all you wanted.

Handwriting, in any useful way. The engine is trained on printed characters, and a handwritten note comes back as confident nonsense rather than an empty page, which is worse.

Tables come out as text in roughly the right reading order, not as a spreadsheet. And this is not a PDF to Word converter. You can search the finished file, copy out of it and quote it, but you cannot edit the sentences, because your pages stay as images with the recognized words sitting behind them.