Drop in an image and get editable text back, without uploading the picture to a server.
The tool uses Tesseract, a well known open source OCR engine compiled to run inside your browser. The first run downloads the engine and the language data, which can take a few seconds. After that the recognition runs locally and the progress bar shows how far along it is.
Results depend on image quality. Straight, high contrast text at a reasonable size is read very accurately, while skewed, blurry or decorative text will contain mistakes. The recognized text appears in an editable box so you can correct it before copying.
The recognition runs in your browser, so the image is not sent to a server for processing. Your browser does fetch the engine files and language data from a public content delivery network the first time.
Use the highest resolution you have, crop to the text, straighten the page, and avoid shadows. Black text on a plain light background gives the best results. Choose the right language before you start.
Not reliably. The engine is designed for printed text. Neat block capitals sometimes work, but cursive usually does not.