How OCR reads Arabic and English on a page
What the reader can and cannot do, and how to check a scan before you trust the text.
OCR turns a picture of letters into characters you can copy. It is a reading pass, not a guarantee. A blurred stamp, a signature, or a table with tiny footnotes will come back wrong even when the headline looks perfect.
A photo of a page
Use Image to Text for a JPG, PNG, or WebP. Choose Arabic, English, or both when the page mixes them. Both is slower and better when one line is Arabic and the next is English. The text box follows the script: Arabic runs right to left, English stays left to right.
Read the result before you paste it into a form. Numbers and names are the usual failures. If a line is nonsense, the photo is the problem: go back to Document Scanner, straighten the page, and try again.
A PDF that is only pictures
OCR a scanned PDF reads the first pages and gives you copyable text. It does not write a hidden text layer back into the PDF. If you need a smaller PDF as well as the text, compress or export images separately. Do not expect the download to be a searchable PDF.
What to keep
- Copy the text, or download it as a text file.
- For several images, switch between the results instead of scrolling one mixed block.
- Keep the original scan. The text is a transcript, and transcripts have mistakes.
Recognition runs in the browser after the language data downloads. The page image is not uploaded to 1CLIK.