Bangla OCR: Image to Bangla Text, in Your Browser

Take a photo of a printed Bangla page, a notice, a book or a screenshot, and get the text back as editable Unicode Bangla. The recognition engine and the Bengali language model load into your browser from this website, and the picture is read on your device, so a private document does not have to be sent to anyone.

OCR for Bangla is genuinely hard, and this tool does not pretend otherwise: it shows how confident the reader was, flags blurry pictures, and gives you the text in a box you can correct. Treat the output as a first draft.

সংক্ষেপে: ছাপা বাংলা লেখার ছবি থেকে টেক্সট বের করুন। ছবি আপলোড হয় না। ফলাফল সবসময় মিলিয়ে দেখুন।

Loading the tool…

Overview

About Bangla OCR: Image to Bangla Text

The engine. Recognition uses Tesseract (compiled to WebAssembly) with the Bengali "best, integer" model published for Tesseract. The engine, the Bengali model (about 1.4 MB) and the English model are served by this site. The Bangla model is downloaded the first time you press the button; nothing from your picture is part of that request.

Before recognition. Optional steps help typical phone photos: converting to grey, evening out uneven lighting, boosting contrast, straightening tilted lines and enlarging small pictures. You can turn each off if it makes things worse for your picture, and rotate the picture in 90° steps when it is sideways.

Language and layout. Choose Bangla, or Bangla + English when a page mixes both scripts. The layout setting tells the engine whether to expect a full page, one block, one column or a single line; a single line or block often works better for headings and captions.

Digits. Bangla digits (০-৯) are recognised where the model sees them. You can keep digits as read, convert them all to Bangla digits, or convert them to English digits afterwards.

Confidence. The percentage is the engine's own estimate and is not accuracy. Words it was unsure about are counted, so you know how much proofreading to expect. A blur check warns you when the picture is too soft to read reliably.

Compatibility

Supported Formats

JPEG, PNG and WebP pictures up to 30 MB, 12,000 pixels on a side and 40 megapixels, up to 12 pictures at a time (decided by the content of the file, not its name). For a PDF page, save the page as an image first (see the PDF to JPG tool). Printed Bangla text works best; handwriting is not supported.

Guide

How to Use Bangla OCR: Image to Bangla Text

Choose one or more pictures (JPEG, PNG or WebP). Rotate them if they are sideways.
Pick Bangla or Bangla + English, leave the picture improvements on, and press Read the text. The first run loads the language model.
Read the result in the box, compare it with the picture, and correct mistakes by typing in the box.
Copy the text or download it as a UTF-8 .txt file.

Highlights

Features

Runs in your browser The picture is read on your device; the engine and Bengali model load from this site.
Bangla and English Recognise Bangla only, or Bangla together with English words.
Picture improvement Grey conversion, even lighting, contrast, straightening and enlarging, each optional.
Honest confidence Shows the engine's confidence, counts uncertain words and warns about blurry pictures.
Bangla or English digits Keep digits as read or convert them to ০-৯ or 0-9.

Applications

Use Cases

Getting text out of a photographed page, notice or form
Digitising a short passage from a printed book for quoting or searching
Reading a screenshot of Bangla text that cannot be selected
Turning a scanned page that has no text layer into editable text
Checking what a document says before deciding to retype it

Good to Know

Limitations

OCR makes mistakes. Bangla has joined letters (conjuncts), small vowel marks and many look-alike shapes. Decorative or very small fonts, low resolution, blur, shadows, curved pages, coloured backgrounds and mixed languages all lower accuracy, and the result can contain wrong, missing or extra letters. Proofread every result.

Handwritten text, tables and complex multi-column layouts are not reliably recognised, and the tool does not keep formatting. Large pictures take time on a slow device; use the Stop button if needed. If the first line is wrong, the picture is probably the cause: a sharp, straight, evenly lit photo helps more than any setting.

This tool runs entirely in your browser — your file is never uploaded to our servers. See our Privacy Policy for full details.

Support

Frequently Asked Questions

Is my picture uploaded?

No. The picture is processed in your browser tab. Only the OCR engine and the Bengali language model are downloaded from this site, and your picture is not part of those requests.

How accurate is it?

It depends on the picture. Clean, large, printed text can come out well; small, blurred, decorative or low-contrast text often does not. The confidence figure is an estimate, not a measured accuracy. Always proofread.

Can it read handwriting?

No. The model is trained on printed text.

The text is gibberish or empty. What can I try?

Rotate the picture so the lines are horizontal, switch the picture improvements on or off, try a different layout (one block or one line), and use a sharper, brighter photo taken straight on.

Can it handle Bangla and English together?

Choose "Bangla + English". English words inside Bangla text are then recognised with the English model too, with the usual risk of mistakes at the boundaries.

My PDF has Bangla but copying gives strange Latin letters. Is that OCR?

No, that is a legacy font (Bijoy or SutonnyMJ). Use the Bijoy PDF to Unicode tool; it reads the real text and converts it, which is more reliable than OCR.

Keep Going

Bangla OCR: Image to Bangla Text (ছবি থেকে বাংলা লেখা)