OCR: recognize text in a PDF or photo
Turn a scanned PDF or a photo of a document into text you can copy, edit and search. Recognition runs on your device, so your file is never uploaded.
Works best on sharp, straight, well-lit scans of printed text. Handwriting is not supported.
Processed on your device. Your files sent to any server: 0 bytes.
How to use
- Add a scanned PDF or a photo of a document.
- Choose the language of the text and press Recognize text (the first run loads the language data from this site, about 2 to 3 MB).
- Copy the text, download it as TXT, or download a searchable PDF.
FAQ
Are my files uploaded anywhere?
No. Recognition runs in your browser. The engine and the language data are served from zenbek.com and kept by your browser, and your file is not sent to any recognition server.
Which languages are supported?
English, Russian and Azerbaijani (Latin script), plus the combinations English + Russian and English + Azerbaijani. Choose the language of your document: a wrong choice lowers accuracy sharply.
Can it read handwriting?
No. The tool is built for printed text. Handwriting gives poor results.
How accurate is it?
On a sharp, straight scan of printed text it is usually good, and the tool shows an approximate confidence figure. Blurry photos, shadows, tiny fonts and skewed pages reduce quality, so always proofread names and numbers.
How large can the file be?
Up to 50 MB. For a PDF, the first 30 pages are processed in one run, so split a longer document first.
What is a searchable PDF?
A copy of your document in which the original page image stays as it is and an invisible text layer is added. You can then search the file and select text in it.
Tips for better results
Scan at 200 to 300 DPI, or take the photo straight above the page in even light, without shadows or glare. Crop away the table or background around the page.
Pick the language that matches the document. For a text with both English and Russian words, use the combined option; it is a little slower but recognizes both alphabets.
Columns, tables and forms can come out in a different order than on the page, so check the result before you use it. Pages with a password must be unlocked first with Unlock PDF. Large pages are scaled down so that the longest side is at most 3200 pixels.
Everything stays on your device, which makes the tool suitable for documents with personal data. After recognition you can tidy the text with the text cleaner.