Clear images give better results
Use a sharp, upright image with readable text and good contrast. Crop away unnecessary borders. PNG, JPEG and WebP files up to 15 MB and 24 megapixels are supported. Handwriting, decorative fonts, columns and blurred photographs may need substantial correction.
English, with optional Nepali
English is enabled by default. Add Nepali for Devanagari text or mixed English and Nepali documents. Review names, numbers and punctuation before using the result. OCR does not preserve the original document layout.
Local processing, downloaded language models
Tesseract.js recognizes text on your device. Starting OCR downloads its worker, WebAssembly engine and selected language models from third-party CDNs (jsDelivr and Project Naptha). These services receive normal connection information, but your image and extracted text are not uploaded. The browser may cache language models for later runs. A connection is needed for uncached assets.