What OCR reads well
Printed text with good contrast: screenshots, scanned letters and invoices, book pages, signs and slides photographed straight on. You get editable text to copy into a document, a search or a translation, instead of retyping it. Paste a screenshot with Ctrl+V or drop a JPG, PNG, WebP or BMP.
Tips for better accuracy
Pick the language of the text: accents and words are recognised far better. Photograph the page flat and square, in even light, with the text filling the frame; crop away margins and backgrounds. Large, sharp characters beat a tiny line in a big photo. Handwriting, stylised logos and text on busy backgrounds give poor results.
How it works and its limits
Recognition uses Tesseract, an open-source OCR engine, compiled to run in your browser. The model of the language you pick (a few MB) downloads on first use and is then cached. Layout is not kept: you get the text line by line, without columns or tables. For a PDF that already contains selectable text, PDF to Text is faster and exact. Images up to 20 MB.