Image to Text (OCR)
Choose a photo, screenshot or scanned PDF and the recognised text appears ready to copy. Recognition runs on your device with the open-source Tesseract engine.
Recognition runs in your browser. The engine and language model are fetched from a CDN on first use; your files are never uploaded.
About this tool
Optical character recognition turns pictures of text into editable text: a photo of a page, a screenshot of an error message, a scanned contract. This tool runs Tesseract, the open-source OCR engine, compiled to WebAssembly so it works inside your browser. The engine and the language model for the language you choose are downloaded from a content delivery network on first use and cached; your images are processed locally and never uploaded.
Accuracy depends on the input: sharp, upright, high-contrast text in a language the model knows reads almost perfectly, while low-resolution photos, handwriting and decorative fonts degrade quickly. For scanned PDFs, every page is rendered and recognised in turn. Choosing the right language matters because the model uses its dictionary to resolve ambiguous characters.
Frequently asked questions
- Are my images uploaded?
- No. Recognition happens in your browser. Only the OCR engine and language data are fetched from a CDN, and they contain nothing of yours.
- Why is the first run slow?
- The engine and language model, a few megabytes, are downloaded once. Later runs reuse them from the cache.
- Does it read handwriting?
- Rarely. Tesseract is trained on printed text; neat block capitals sometimes work, cursive does not.
- How do I improve accuracy?
- Use a sharp image with the text upright and at least around 30 pixels tall, crop away clutter, and pick the correct language.