Read the text out of a photo, screenshot or scan and turn it into something you can edit and copy. Twelve languages including Arabic, running entirely on your own device.
Drop an image containing text
or click to browse · a photo, screenshot or scan · JPG, PNG, WEBP
English, French, Arabic, Spanish, German, Italian, Portuguese, Dutch, Turkish and Russian, plus combined modes for documents that mix two languages.
Greyscale conversion, contrast boosting and upscaling of small images run before recognition, which makes a substantial difference on photographs taken with a phone.
Most free OCR services upload your file. Here the recognition engine runs in your browser, so a payslip, ID or contract never leaves your device.
Step by step
A screenshot, a photo of a page, or a scan. Sharper and straighter images give better results.
This matters more than anything else. Recognition uses a language model, so selecting Arabic for Arabic text roughly doubles accuracy compared with leaving it on English.
Press extract. The first run downloads the language model, which takes a few seconds; after that it is cached by your browser.
The result appears in an editable box. OCR is never perfect, so proofread it, then copy the text or download it as a .txt file.
FAQ
Optical character recognition converts pictures of text into actual text characters. It analyses shapes in the image, matches them against a trained model of a language, and outputs the letters it believes it sees.
On a clean screenshot or a flat scan, accuracy is usually very high. On a photograph taken at an angle in poor light, it drops sharply. The confidence figure shown after recognition is a useful guide: below about 70 percent, expect to do real editing.
Photograph the page straight on rather than at an angle, use even lighting without shadows or glare, fill the frame with the text, and make sure the image is in focus. These matter far more than any setting in the tool.
Yes. Arabic is one of the supported languages, and there is a combined English and Arabic mode for mixed documents. Arabic is harder for OCR than Latin scripts because letters connect and change shape by position, so expect to proofread more carefully.
Generally no. The engine is trained on printed type. Very neat block capitals sometimes work, but cursive handwriting is beyond what this kind of recognition can do reliably.
The language model has to be downloaded, which is a few megabytes. Your browser caches it, so subsequent extractions in the same language start immediately.
No, and this is the main reason to use this tool rather than a typical free OCR site. The recognition engine runs inside your browser, so the image and the extracted text stay on your device.
This tool performs OCR online to extract text from an image, whether that is a screenshot, a photograph of a printed page, or a scanned document. The result appears in an editable box you can correct, then copy or download as a text file.
Twelve language options are available, including Arabic, French, Spanish, German, Turkish and Russian, plus combined modes for documents that mix two languages. Choosing the correct language is the single most important setting: recognition works by matching shapes against a trained model, so the wrong model produces markedly worse results.
Before recognition runs, the image is prepared automatically. Converting to greyscale removes colour noise, boosting contrast separates letters from the background more cleanly, and small images are upscaled because the engine performs better with more pixels per character. On phone photographs these steps make a real difference.
What matters most, though, is the source image. Photographing a page straight on rather than at an angle, with even lighting and no glare, improves accuracy more than any software setting. The confidence percentage reported after each run tells you how much proofreading to expect: high confidence usually needs a quick scan, low confidence needs a careful read.
Unlike most free OCR services, nothing is uploaded. The recognition engine runs inside your browser, which matters when the document is a payslip, an identity paper or a contract. Related tools include the PDF to Word converter for PDFs that already contain a text layer, and the photo editor for sharpening a difficult image first.