Make a scanned PDF searchable with OCR - locally in the browser, with an invisible text layer under the original pages, no upload.
Running locally on your device ...
0%
Your files never left your device
No. Everything runs in your browser - your file never leaves your device. How this is verifiable
A scanned PDF is essentially a stack of images: you see the text but cannot select, search or copy it. OCR (optical character recognition) reads the letters out of the image and places them as an invisible text layer exactly over the pages. The look stays identical, but the PDF becomes searchable and the text selectable - exactly what offices, archives and full-text search expect.
Choose the output: a searchable PDF (original pages plus text layer) or plain text as TXT. Set the document language from the list - over 100 languages from German and English through Russian, Arabic and Hindi to Japanese, Korean and Chinese, each in its own name - so the characters are recognised correctly. Everything runs in the browser via WebAssembly: the same open-source Tesseract engine as the "Image to text" tool, applied page by page across the whole PDF. On first use the browser downloads the engine and the chosen language model once (a few MB, same-origin); after that it works offline.
The tool is meant for scanned or image-based PDFs. Recognition depends heavily on scan quality: sharp, straight, high-contrast pages of printed text are recognised very well, skewed or blurry ones much worse; handwriting not reliably. In the searchable PDF each page is re-embedded as an image - so a PDF that already contains real text does not need OCR. Glance over the result for numbers and proper names.
| Input formats | |
|---|---|
| Output format | |
| Batch processing | Yes |
| Processing | Locally in your browser (WebAssembly) |
| File upload | None |
Limitations: For SCANNED / image PDFs: makes them searchable via OCR (invisible text layer) or returns plain text. Recognises PRINTED text in over 100 languages and many scripts; handwriting not reliably. Note: the invisible text layer in the searchable PDF uses a Latin-1 font - non-Latin characters (e.g. CJK, Arabic, Cyrillic) are not embedded there; for those scripts the "Text only (TXT)" output returns the fully recognised text. In the searchable PDF each page is re-embedded as an image, so the layout is preserved but the file can grow; a PDF that already has real text does not need OCR. Recognition is not error-free - check numbers and proper names in particular. Engine + chosen language model are downloaded once (a few MB, same-origin), then offline-capable. Large PDFs take some time and memory depending on page count and device.
No. Rendering and recognition run entirely locally in the browser (WebAssembly); the PDF never leaves your device - the decisive advantage for contracts and official mail in particular.
The page looks like the scan, but underneath sits an invisible text layer. That lets you search the PDF, select and copy text - while the visual impression stays the same.
Yes. Each page is embedded as an image of the original page, with the invisible text layer aligned over it. The look does not change, though the file can grow a little.
Other languages yes: over 100 are available (including Arabic, Russian, Hindi, Japanese, Korean, Chinese). Handwriting is not recognised dependably, only printed text. For non-Latin scripts use the "Text only (TXT)" output, since the searchable-PDF text layer is limited to Latin-1.