Image to Text (OCR)

OCR (Optical Character Recognition) reads the letters in a picture and turns them into real, selectable text you can copy and edit. Use it to pull the text out of a screenshot, copy a receipt or an invoice, lift a quote from a photo of a page, digitise a scanned document, or grab text from a whiteboard photo — anything where the words are locked inside an image. This tool recognises the text and gives you back plain text you can copy or download.

No — the recognition runs entirely in your browser using a self-hosted OCR engine (Tesseract compiled to WebAssembly). Your image is read and processed on your own device and is never sent to a server, logged or stored. The first time you run it, the English language model is downloaded once from this site and then cached, after which the tool works offline. That makes it safe for receipts, documents and screenshots you would rather not upload anywhere.

Accuracy depends heavily on the image. Clear, high-contrast text — a screenshot, a flat scan, printed black text on white — is recognised very well. Photos taken at an angle, low-resolution images, busy backgrounds, unusual fonts, and especially handwriting are much harder and may come out with mistakes or gaps. For the best result, use the sharpest, most straight-on and highest-contrast version of the image you have, and crop away anything that is not text before running it.

The OCR tool accepts JPG, PNG, WebP, GIF and BMP — any image your browser can decode. For an iPhone HEIC photo, convert it to JPG first with the HEIC to JPG tool, then run OCR. Text recognition currently uses the English model, which also handles the digits and common punctuation found in receipts and documents; it is not tuned for other languages or scripts. You get the extracted text back as selectable text you can copy, or download as a .txt file.

Handwriting only works unreliably. This OCR engine is trained on printed and typed text, so neat, consistent handwriting sometimes comes through partially, but cursive and casual handwriting usually do not. If your goal is handwriting, treat any result as a rough starting point that will need heavy correction rather than a finished transcription. For printed or typed text, the results are much more dependable.