Why no searchable PDF yet

A scanned PDF is usually just pictures of pages. A searchable PDF has the same pictures plus an invisible text layer lined up under the words, so find, select, and copy work while the page still looks like the scan.

To check a file, open it in any PDF reader, press Ctrl+F (Cmd+F on a Mac), and search for a word you can see on the page. If the reader finds nothing, or you cannot select a line of text with the cursor, the file has no text layer yet.

We tested this with a two-page PDF made from two page images, the shape a scanner produces. The pdftotext command line tool found no words in it at all. You can download that test PDF and use it to try the options below.

These three tools are free, and each runs on your own computer, so the PDF does not leave it. All three use Tesseract, the same open-source OCR engine Screenshot2Text runs in the browser.

OCRmyPDF is a command line tool that adds a text layer to a scanned PDF and keeps the original page images. The basic command is ocrmypdf -l eng input.pdf output.pdf. Its install guide covers apt on Debian and Ubuntu, Homebrew on macOS, and Windows. By default it writes PDF/A, an archival flavor of PDF. Add --output-type pdf for a regular PDF, and --deskew to straighten crooked pages.

NAPS2 is a free, open-source scanning app for Windows, Mac, and Linux with a normal window and buttons. Click OCR in the toolbar, download a language, and tick Make PDFs searchable using OCR. Its OCR guide says it can also add OCR to PDFs you import, and it leaves pages that already have text alone.

The Tesseract command line makes searchable PDFs from page images. tesseract page.png page -l eng pdf writes page.pdf with the image and a text layer. For a whole document, put one image file name per line in a text file and run tesseract list.txt out -l eng pdf, which writes a single multi-page out.pdf, as the Tesseract manual describes. We tried it with two page images and got a two-page searchable PDF. It reads images, not PDFs, so a scanned PDF needs its pages saved as images first. OCRmyPDF does that step for you.

Screenshot2Text does not write a text layer into the PDF. It gives you the words. Drop the PDF on PDF to text or the homepage, and each page is drawn and read in your browser tab. Nothing is uploaded and no account is needed. The text lands in an editable box, with a header line before each page, and you can copy it or download a .txt file.

On the same two-page test PDF it took 1.63 seconds for both pages (median of three runs, local build of the current code in headless Chromium), and the text had one mistake: sign in came back as signin. The PDF to text page shows the full output.

That is often all a searchable PDF was for: quoting a clause in an email, pulling a figure into a spreadsheet, or checking what a scan says. If the next step is a paste, you do not need a new PDF.

Fix the text first. A text layer is only as good as the OCR under it. If signin stays signin, a search for sign in will miss it, in a searchable PDF or in a text file. Skim names, numbers, and dates in the box before you save.

Save the .txt next to the PDF with the same name, such as lease-2026.pdf and lease-2026.txt. Most desktop search tools index plain text files, so a search for a word from the scan will find the .txt, and the .txt points you to the PDF.

Paste the text into your notes app or a document if that is where you search anyway. You can also paste it into a word processor and export a new PDF, but that is a fresh document made of text, not the original scan with a layer under it. If someone asked for the original pages with searchable text, use one of the tools above.

Writing a trustworthy text layer means placing every word under the right ink, handling columns, and not damaging the original pages. The tools above have years of work behind that. Doing it inside a browser tab, without uploading the file, would add weight and new ways to fail, and a layer that looks searchable but selects the wrong words is worse than none.

So Screenshot2Text stays with what it does well: reading the pages in your browser and handing you editable text. If that changes, this page will change with it.