Read the text out of a photo or scan - locally in the browser via OCR, with 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
OCR (optical character recognition) turns the picture of a text back into real, selectable text. You photograph or scan an invoice, a letter or a book page and get the content as a TXT file you can search, copy and process further. The recognition model is the open-source Tesseract engine, running here entirely as WebAssembly in the browser - the same approach as the other tools.
Pick the document language from the list - over 100 languages are available, from German, English and the European languages through Russian, Arabic, Hebrew and Hindi to Japanese, Korean and Chinese. Each language is shown in its own name, so you spot it at once. The right choice makes the recogniser read the letters, accents and special characters of that script correctly. On first use the browser downloads the recognition engine and exactly the chosen language model once from this site (a few megabytes, same-origin, no foreign server); after that the tool works offline too. Drop several images at once - each becomes its own TXT.
Recognition quality depends almost entirely on the source: a sharp, straight, high-contrast photo of printed text is recognised very well; a skewed, blurry or dim image much worse. Printed type works reliably, handwriting does not. Columns, tables and unusual layouts come out as running text - the order is usually right, but the visual structure is lost. Always glance over the result, especially for numbers.
| Input formats | JPG, JPEG, PNG, WEBP, BMP |
|---|---|
| Output format | TXT |
| Batch processing | Yes |
| Processing | Locally in your browser (WebAssembly) |
| File upload | None |
Limitations: Recognises PRINTED text from clear images (JPG, PNG, WebP, BMP) as TXT; handwriting is not recognised reliably. Over 100 languages and many scripts (Latin, Cyrillic, Greek, Arabic, Hebrew, Indic scripts, CJK and more) - pick the one matching your document. Columns/tables/layout are not preserved as structure, only as running text. Recognition is not error-free - check the result, especially numbers and proper names. On first use the engine plus the chosen language model is downloaded once (a few MB, same-origin); offline-capable afterwards. For scanned PDFs there is a dedicated "PDF OCR" tool (searchable or plain text).
No. Text recognition runs entirely locally in the browser (WebAssembly); the image never leaves your device - the decisive advantage for sensitive documents in particular.
Over 100 languages and many scripts - besides German and English also French, Spanish, Russian, Arabic, Hebrew, Hindi, Japanese, Korean, Chinese and more. Pick the matching language from the list so special characters and letters are recognised correctly.
No, reliably only with printed text. The engine does not recognise handwriting dependably.
OCR depends on image quality: sharp, straight and high-contrast is recognised very well, skewed/blurry/dark much worse. Always double-check numbers and proper names.
Yes, there is a dedicated "PDF OCR" tool for that: it makes a scanned PDF searchable directly or returns plain text - with no detour via single images.