Image to Text (OCR)

The tool uses Tesseract.js, a WebAssembly build of the open-source Tesseract OCR engine. Before recognition, the image is prepared to help the engine: small images are enlarged so their width approaches 2500 pixels (never more than three times the original size), and the colours are converted to high-contrast greyscale. Tesseract then finds lines of text and converts them to characters. The page layout is detected automatically.

Recognition starts as soon as you choose a file, and a progress indicator shows each stage.

• Works well: screenshots, printed documents, receipts and book pages photographed flat, with dark text on a light background.
• Works poorly: handwriting, decorative fonts, text on busy photos, very small text, curved or strongly angled pages, and light text on dark backgrounds.
• Language: only English is loaded. Other languages that use the Latin alphabet are partly recognised but accents and special letters are often wrong; other scripts are not recognised.
• Layout: the output is plain text. Columns, tables and formatting are not preserved, and multi-column pages may come out interleaved.

• Crop to just the text you need with the Crop tool; less background means fewer stray characters.
• Rotate pages so lines are horizontal with the Rotate tool.
• Use the sharpest original you have. A screenshot beats a photo of a screen.
• Always proofread. OCR commonly confuses similar shapes such as 0 and O, 1 and l, or rn and m.

The first time you use the tool, your browser downloads the OCR engine and the English language data from the jsDelivr content network. These are program files only; your image is not sent with them. Tesseract stores the language data in your browser's storage, so later runs start faster.

Recognition itself happens on your device. The extracted text appears on the page and is copied only when you press copy. Nothing is uploaded to Image York.