How to extract text from an image or a scanned PDF

How to do it

1. Drop the image or PDF into the tool above. Photos, screenshots, scans and iPhone HEIC photos all work.
2. Choose the language the text is written in. This matters more than any other setting.
3. Let it read. The first run for a language downloads that language’s data; later runs start straight away.
4. Copy the text, or download it as a Word document with headings, lists and tables kept.

For a stack of pages or a folder of screenshots, use batch image to text, which reads them all in one pass and gives you one file or one file per page. To read text in a language you do not speak, image translator recognises it and translates it in one step.

Why the language matters so much

Text recognition does not just match shapes to letters. It uses a model of the language to decide between readings that look alike, such as “rn” and “m”, or “0” and “O”. With the wrong language chosen, perfectly clear text turns into near misses. Choose the language on the page, and for a page that mixes two, choose the one most of the text is in.

The tool covers 102 languages, from English, Spanish and Indonesian to Arabic, Chinese, Japanese, Korean and Javanese.

Getting an accurate result

Most recognition errors are really photo errors. Before blaming the tool, check these:

• Flat and straight. A page photographed at an angle, or curving at the spine of a book, is the biggest single cause of mistakes.
• Even light. Shadows across a page break words apart. Daylight from a window beats a lamp overhead.
• Enough pixels per letter. Fill the frame with the text. A whole page photographed from across a desk gives each letter too few pixels.
• Sharp focus. Tap the text on your phone screen to focus before you shoot.
• Crop to the text. Cutting away the table and background removes noise the tool would otherwise try to read.

After recognition, check numbers by eye. Amounts, account numbers and dates are where a single misread character matters most.

Where it struggles

Being clear about this saves time:

• Handwriting. It is built for print.
• Very low resolution. Text a few pixels high cannot be recovered.
• Complex layouts. Magazines with text wrapped around images may come out in the wrong reading order.
• Stylised fonts. Decorative and script typefaces are read less reliably than plain ones.

For difficult material, a commercial recognition engine may read more of it correctly, and it is worth comparing on a sample page.

Keeping the image private

The images people most need text from are often sensitive: ID cards, bank statements, contracts, letters from a doctor, screenshots of internal systems. Many OCR sites upload them first. Here the recognition happens on your device, so the image never leaves it.

Reembun. (2026, September 11). How to extract text from an image or a scanned PDF. https://reembun.com/guides/extract-text-from-image