CornerTools

Searchable PDF (OCR)

Make a scanned PDF searchable, or copy all its text out, in English and Hindi, without uploading it.

Drop PDFs here, or

Your files never leave your device

Processed in this browser · up to 50 at a time

Language

The text reader downloads the first time a page needs it, 11.97 MB for English; its models are kept for later visits.

Clear, straight scans read best. Handwriting is not read reliably.

A scanned PDF is pictures of pages: you cannot search it, select a line or copy a figure out of it. This tool reads the text on those pages (optical character recognition, OCR) and either adds it to the PDF as an invisible layer, so the PDF looks the same but can be searched and copied from, or saves it as a plain text file. It reads English and Hindi, and also fixes PDFs whose Hindi text copies out as garbled letters. Everything runs on your device; nothing is uploaded.

How to use it

Add one or more PDFs; a PDF with a password asks for it. Then choose:

  • Searchable PDF or Text (.txt). The searchable copy is named as in scan-ocr.pdf, the text file as in scan.txt.
  • Language: English, Hindi, or both for pages that mix them. Choosing only the language a document is in reads it more accurately.
  • Pages: such as 1-3, 5; empty for every page.
  • Read every page: reads pages that already have text too (see below).

The text reader downloads from this site the first time a page needs it: the engine (up to 4.7 MB), a model that tells which way up a page is (4.32 MB), and each language's model (English 2.95 MB, Hindi 1.39 MB): about 11.97 MB for English. Your browser keeps the models, so later visits download much less.

How it works

Each page is first checked: drawn at screen resolution, it is compared with the text it already holds. A page is read when it has no text of its own; when it is a scan, one picture covering the page, whose text covers little of what is printed (a scanner app's small footer of text does not count); or when the text it has is garbled. Pages of text with photos on them, and pages whose text is fine, are left alone, and a blank page is skipped.

A page to read is drawn at 300 DPI. A model works out whether it is sideways or upside down, and the page is read turned upright; the page in the PDF keeps its own orientation. The text is read by Tesseract, the open-source OCR engine. On a scan that already has a little text of its own, that text is left out of the reading, so it is not found twice.

In the searchable PDF, each word is placed over the word it was read from, in invisible text, so search highlights the right place and selecting a line selects the words under it. The pages themselves are not drawn again: images, form fields and links stay exactly as they were, and the file grows only by the text. In the text file, each page gives its own text where it had good text and the text read from it otherwise, with a blank line between pages.

Frequently asked questions

Many Hindi documents, government ones especially, were typed in older fonts such as Kruti Dev and DevLys, which draw Devanagari through the codes of Latin letters. The page looks right, but the text behind it is those letters: भारत copies out as something like Hkkjr. This tool recognises those fonts and reads such pages from how they look. In a searchable PDF the old text stays under the new, so search and copy find both; the text file has only the new.
Clean, straight printed pages read almost word for word. Accuracy drops with blur, low resolution, faint or crumpled paper, unusual fonts and tables, and handwriting is not read reliably. Check names and figures before you rely on them. Words read wrongly in a searchable PDF are wrong only in its invisible layer: the page still shows the original.
No. The text layer is invisible and does not print. A digitally signed PDF, such as an e-Aadhaar, will no longer verify after any change, this one included; the tool says so when you add one.
No. Your browser reads, reads and saves the files on your device; they are not sent to CornerTools or anyone else, and neither is a password you type. To check, open the browser's developer tools at the Network tab and add a PDF: no request carries it. The PDF engine downloads from this site the first time you use it: a download to your browser, not an upload of your file.

See also

  • JPG to PDF — to make photos of pages into a PDF to read.
  • Compress PDF — to make a large scan small enough to upload.
  • PDF Editor — to add text, signatures and images to the pages.