The reading desk · पढ़ने की मेज़
Image to text — Hindi, Bangla, Tamil & more
Recognise Hindi, Bangla, Tamil and other Indian-language text from images or scanned PDF pages on your device. Paste, upload or photograph text, proofread and download Word text.
Source
Choose a page to read
Specimen
हिन्दी भारत
Check ि placement and conjuncts such as क्ष; clear Devanagari print works best.
Images are processed on this device and never uploaded. The first run downloads a language model from jsDelivr; image content is never sent there.
Result
Editable Unicode text
01 · How it works
Hindi: recognition and proofreading
हिन्दी भारत
Check ि placement and conjuncts such as क्ष; clear Devanagari print works best.
Choose the matching language model, upload or paste a clear image, or use the mobile camera input. Keep the page straight and well lit. Recognition is intended for printed text; handwriting, ornate fonts and low resolution may need extensive correction. Large images are reduced to a maximum 3000-pixel edge; PDF pages to 2400 pixels.
Reference
| Language | Model / sample |
|---|---|
| Hindi | hin — हिन्दी भारत |
| Marathi | mar — मराठी महाराष्ट्र |
| Nepali | nep — नेपाली नेपाल |
| Sanskrit | san — संस्कृतम् नमस्ते |
| Bengali / Bangla | ben — বাংলা বাংলাদেশ |
| Assamese | asm — অসমীয়া ৰ ৱ |
| Gujarati | guj — ગુજરાતી ગુજરાત |
| Punjabi / Gurmukhi | pan — ਪੰਜਾਬੀ ਪੰਜਾਬ |
| Odia | ori — ଓଡ଼ିଆ ଓଡ଼ିଶା |
| Tamil | tam — தமிழ் வணக்கம் |
| Telugu | tel — తెలుగు నమస్కారం |
| Kannada | kan — ಕನ್ನಡ ನಮಸ್ಕಾರ |
| Malayalam | mal — മലയാളം നമസ്കാരം |
| Urdu | urd — اردو سلام |
| English | eng — Hello India |
02 · FAQ
Questions about reading text from files
Are my images uploaded?
No. Recognition runs on your device. The recognition file for your language downloads from a public file host (jsDelivr) the first time you use it; that host sees a normal download request, never your image.
Is the confidence score accuracy?
No. It is the engine’s estimate. Proofread names, numbers, vowel marks and conjuncts.
Can I recognise a scanned PDF?
Yes. PDF.js renders pages one at a time, then the OCR worker recognises each image. Output is editable text and Word paragraphs, not the PDF layout.