How to make a scanned PDF searchable with OCR: a step-by-step guide

Drop the PDFs anywhere on the page, or press Add files. Each file is checked on arrival for pages without text - the file list says how many it has, and the pages to be read are highlighted. A file that already has text says which language it is in.

For a scan with no text at all, the tool reads its first page quickly to guess the language ("first page: English" above the thumbnails), and the language chips follow it.

• Detect the language (on by default) - every page is judged from its own words, and a document in two languages is read in both. The chips show the guess for the selected file; click one to override it.
• Add a language - pick from the 40 languages the tool reads, Latin, Cyrillic, Greek, Arabic, Hebrew and Chinese, Japanese and Korean scripts included.
• Turn detection off to read every page with exactly the languages you chose.

The first run downloads the recognition engine (about 4 MB) and the language data (1-2.5 MB each) - from our own site, never a third party. The browser keeps them, so the next run starts at once.

Which pages is Without text by default: only the pages that have no text of their own are read - a page made on a computer already has its text, and a second layer would double every search hit. Choose Every page to read a poorly recognised scan again; its old layer stays underneath.

Press Recognise text. The line above the thumbnails and the button show the progress page by page. When the pages are read, the searchable PDF is saved to your device at once, and the file list shows how many words were found and in which language.

Open the result in any PDF reader and search for a word: it is found and highlighted on the scan. Words the engine was unsure of are still placed - a search wants them - and the result counts them as uncertain.

• Scan at 300 dpi if you can. Under 150 dpi the tool warns that it expects errors - it can only read what the picture holds.
• A crooked scan reads better straightened: Straighten PDF (Deskew) levels it first.
• Pages scanned sideways or upside down? Auto-rotate PDF puts them upright.

Without an account, and with a free one, you can recognise two files in a run, each up to 15 MB. With Pro there is no limit on the number or the size, and you can save your languages and page choice as a preset. Every page is read on your own computer, so a long document takes as long as your machine needs.

• Edit the recognised text in Word? PDF to Word.
• Only the text, as a .txt? PDF to Text.
• A scan that is too big to send? Compress PDF.