OCR PDF
Adds an invisible text layer to scanned PDFs and photos so you can search, select and copy the text. 26 languages. The page images aren’t changed.
What OCR PDF does
OCR PDF makes a scanned PDF searchable. Text recognition runs in your browser and adds an invisible text layer on top of each page, so you can search, select and copy the text while every page still looks exactly like the scan.
It also works on photos and scans saved as images (JPG, PNG, WebP or BMP), which become one searchable PDF.
How to make a scanned PDF searchable
- Drop a scanned PDF, or several images, on the box above.
- Choose the Document language and, if the text mixes two, a Second language. Pick an Accuracy level.
- Select Make searchable. When it's done, download the PDF; the summary shows how many pages were recognized.
Settings
| Document language | 26 languages; see the list below. |
|---|---|
| Second language | Optional, for documents that mix two languages. |
| Accuracy | Fast reads pages at 150 dpi, Balanced at 220 dpi (default), Best at 300 dpi for small print. |
| Skip pages with text | On by default for PDFs: pages that already have selectable text are left alone. |
| Languages | English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Turkish, Russian, Ukrainian, Arabic, Hebrew, Hindi, Bengali, Tamil, Telugu, Marathi, Urdu, Chinese (Simplified), Chinese (Traditional), Japanese, Korean, Vietnamese, Indonesian, Thai. |
Limits
- Handwriting is recognized poorly or not at all.
- The first run is slower while the engine and language model load.
- Recognition uses your device's processor, so long scans take longer on older devices.
Where your file goes
Recognition runs in this browser tab with Tesseract compiled to WebAssembly. Your scan isn't uploaded. The first run loads the OCR engine and the English model from this site, about 7 MB. The other 25 languages download their model, up to 3 MB each, from the jsDelivr CDN the first time you use them. Your browser caches these files, and nothing about your document is part of the download.
If you're signed in, the result panel also offers Save to My documents. Only that button uploads the file, to your account's storage.
Questions
Does OCR change how my PDF looks?
No. The text layer is invisible and sits on top of the original page images, which aren't modified.
How accurate is it?
Clean printed text at Balanced or Best is usually recognized well; the result panel shows the average confidence. Low-resolution scans, unusual fonts and handwriting reduce accuracy. Search for a few words in the result to check it.
Can I OCR a photo of a document?
Yes. Drop JPG, PNG, WebP or BMP images instead of a PDF. Each image becomes a searchable page in one PDF.
Why is the first run slower?
The OCR engine and the language model download once and are then cached by your browser: English from this site, other languages from the jsDelivr CDN. Later runs start straight away. Your document isn't part of that download.