Which script, and is the source a photo or a PDF? This hub does not run recognition. Each language page loads its own model, and a mixed page uses the bilingual URL. A digital PDF that already has selectable text belongs on PDF to TXT.
Optical character recognition guesses letters from pixels. It does not understand that a line is an address. Proofread numbers. A digital PDF that already has a text layer usually needs PDF to TXT, not OCR.
English, Arabic, and bilingual each load different models in your browser (Tesseract-style). That is why they are separate URLs: the language pack is part of the job. Mixed Civil ID photos belong on the English + Arabic page.
Printed, well-lit, square-on pages work. Handwriting, calligraphy, and photos of projector screens usually do not. Guides explain how OCR works and how to prep a photo.