You have a photo of a notice board, a screenshot of a message, a scanned page from a book or a picture of a form, and you need the words as text. Typing it all again is slow and invites mistakes. Optical character recognition, usually called OCR, reads the letters in the picture and turns them into text you can copy, edit and search. This free image to text converter does that in your browser, with support for Bangla as well as English.

How to use the image to text converter

1. Click the upload box and choose a JPG, PNG or WebP picture. A preview appears so you can check that you picked the right file.
2. Select the language in the image: English, Bangla, Bangla plus English, Hindi or Arabic. Choosing the right language is the most important setting, because the engine uses a different set of shapes for each script.
3. Press Extract Text. A progress message shows what the engine is doing.
4. When it finishes, read the text in the box and correct any mistakes. Then use Copy Text or Download .txt.

The very first run for a language downloads its recognition data, which may be several megabytes. A slow connection can make this take a while. Later runs on the same device are usually faster because the browser can keep the data.

How it works

The tool uses Tesseract, a well-known open-source OCR engine originally developed at Hewlett-Packard and later maintained by Google and the community. A JavaScript and WebAssembly version runs inside your browser. Your image is processed on your own device. The only network traffic is the download of the engine and the language files, not the upload of your picture.

The engine finds lines of text, splits them into words and characters, and compares the shapes with what it learned during training. It reports a confidence number, shown under the result, which is a rough sign of how sure the engine is. A low number tells you to proofread more carefully.

Bangla OCR: what to expect

Bangla is harder for OCR than English. The script has joined letters, many conjuncts, vowel signs that sit above, below, before and after a letter, and a continuous top line. Tesseract has a Bangla model, and with a clean printed page the result is useful. But you should expect more mistakes than in English, especially:

• conjuncts such as ক্ষ, ন্ত and স্ট, which may break into separate letters;
• small vowel signs and the chandrabindu, which can be lost or replaced;
• similar letters such as ব and র, or ণ and ন, in small or blurry type;
• decorative newspaper headlines and old typefaces.

Handwriting is not supported by this kind of engine. It is designed for printed text. If your picture contains handwriting, the output will be mostly wrong.

How to get the best results

• Use a sharp, straight picture. Hold the phone parallel to the page and avoid shadows. A tilted photo makes the lines curve and reduces accuracy.
• Light and contrast matter. Dark text on a plain light background works best. Yellowed paper, watermarks and coloured backgrounds reduce accuracy.
• Use enough resolution. A photo where the letters are at least about 20 to 30 pixels tall works well. Very small screenshots can be enlarged first with the Image Resizer.
• Crop to the text. Remove pictures, borders and columns you do not need so the engine does not try to read decorations.
• Pick the right language. For a page with both Bangla and English, choose "Bangla + English". It is slower, but it reads both scripts.
• One column at a time. For a newspaper with several columns, crop each column separately and run them one after another.

What OCR is useful for

Screenshots and chats. Copy the text from a picture instead of retyping it.

Notices and circulars. Turn a photographed job circular or school notice into text you can search and share.

Books and printed documents. Pull out a quotation or a paragraph from a printed page, with permission and while respecting copyright.

Receipts and forms. Capture figures from a bill so you can paste them into a spreadsheet.

Learning. Collect text from lecture slides and handouts for notes.

After you have the text

Use the Word Counter to check the length, the Text Compare tool to see what changed against another version, or the Remove Duplicate Lines tool to clean a list. If your source document is an old Bijoy file instead of a picture, use the Bijoy to Unicode Converter. If you want to remove hidden data from the original photo before sharing it, try the Free Metadata Cleaner.

Privacy and limits

• The picture stays in your browser. It is not uploaded to our site or to a third-party OCR service.
• The page needs an internet connection the first time to load the OCR engine and language data. If your network blocks the public code delivery network, the tool shows an error message instead of a result.
• Large pictures are slow on older phones. If the browser freezes, resize the picture or crop it.
• PDF files are not accepted here. Convert a page to an image first, for example with the PDF to JPG tool.
• The output has no formatting. You get plain text; tables and columns are not rebuilt.

Frequently asked questions