How to extract text from an image
OCR (Optical Character Recognition) reads the shapes of letters in an image and turns them into actual, selectable, editable text — useful for a photographed notice, a textbook page you want to quote from, or a WhatsApp screenshot with a phone number you'd rather not retype by hand.
What OCR can and can't do
OCR works by recognising character shapes, which means it's fundamentally a pattern-matching problem — it does well on clear, printed text and struggles more with handwriting, unusual fonts, or text at an angle. Hindi (Devanagari) text adds its own complexity because conjunct characters and matras can look visually similar to each other, so accuracy on Hindi text tends to be somewhat more sensitive to photo quality than English text is.
Step-by-step
- Open the Image to Text (OCR) tool.
- Upload a photo or screenshot containing text — Hindi, English, Tamil, Marathi or Gujarati are all supported.
- The tool processes the image and shows the recognised text.
- Copy the text directly — no download step required.
What actually improves accuracy
- Flat, straight-on framing — a photo taken at an angle distorts letter shapes and hurts recognition more than almost anything else.
- Even lighting — a shadow across half the text creates a false edge that can be misread as part of a letter.
- Focus — a slightly blurry photo that looks fine to a human eye can genuinely confuse character recognition, since it relies on precise edges.
- Contrast — dark text on a light background (the normal case) works far better than light text on a dark background or a low-contrast photocopy.
If your PDF needs the same treatment
If you're working from a scanned PDF rather than a single image, the same underlying idea applies through the PDF to Text tool — though that tool works instantly only when the PDF already has a real text layer; a scanned image-only PDF needs the OCR tool applied to its pages instead.
Why OCR accuracy depends on the photo more than the tool
Text recognition works by identifying the shape of each character, so anything that distorts those shapes — a blurry photo, an angled shot of a page, low contrast between text and background, or a crumpled document — reduces accuracy regardless of which OCR tool you use. A sharp, straight-on, well-lit photo of a printed page will usually extract nearly perfectly, while the same tool struggles with a handwritten note or a photo taken at a steep angle under poor lighting, since both distort or obscure the character shapes the tool relies on.
If accuracy matters for a specific document, it's often faster to retake the photo under better light and a flatter angle than to accept a poor extraction and manually correct a long list of errors afterward.
Frequently asked questions
Can this read handwriting?
OCR is tuned for printed text and works much less reliably on handwriting — for handwritten notes, accuracy will vary a lot depending on how neat the writing is.
Does my photo get uploaded anywhere for processing?
No — text recognition runs entirely inside your browser tab, on your own device.
Can it recognise a mix of Hindi and English in the same photo?
Yes, mixed-language text in the same image is supported, though very small font sizes can still hurt accuracy in either language.
Why does text extraction work well on some photos but not others?
OCR relies on recognizing character shapes, so blur, an angled shot, poor lighting or low contrast reduces accuracy regardless of the tool — a sharp, straight-on, well-lit photo usually works best.