What OCR actually does
A photo is just coloured dots. OCR analyses those dots, finds shapes that look like letters, and rebuilds them into text. It works best on printed text with good contrast — dark letters on a light background, photographed straight on.
Why the language setting matters
Each language has its own alphabet and letter patterns. Albanian uses ë and ç, German uses ü and ö, Russian uses a different script entirely. Telling the tool which language to expect dramatically improves accuracy on accented characters.
Getting better results
Photograph the page flat rather than at an angle. Use even lighting with no shadow across the text. Crop out anything that is not text. Handwriting rarely works well — OCR is built for printed characters.
Image or PDF?
If your document is a PDF created on a computer, the text is already inside the file and can be extracted directly — no OCR needed, and the result is perfect. Use OCR only when the text exists purely as pixels, such as a scan or a photo.