8 August 2026
Bangla PDF text won't copy properly — why, and what actually works
Copying Bangla out of a PDF often gives broken letters or nothing at all. The three causes, and the fix for each.
You select Bangla text in a PDF, copy it, paste it — and get scrambled letters, boxes, or an empty clipboard. This has three distinct causes, and the fix depends on which one you have.
Cause 1: it isn't text, it's a picture
Most Bangla PDFs in circulation — circulars, notices, scanned certificates, old books — are photographs of paper wrapped in a PDF. There is no text to copy.
How to tell: try to select a single word. If you can't place a cursor in it, or your selection covers a whole rectangular block at once, it's an image.
The fix: OCR. Our OCR tool reads Bangla in your browser using Tesseract's Bengali model, so certificates and NID scans never leave your device. Expect very good results on a clean 300 DPI scan, and more mistakes around conjunct characters (যুক্তাক্ষর) when the scan is a photocopy of a photocopy.
Cause 2: the PDF has no character map
A PDF can contain real text and still refuse to give it up. Fonts inside a PDF are often subset and embedded with a mapping from glyph shapes back to Unicode characters — the ToUnicode map. When the producing software omits it, the page renders perfectly but the letters have no identity. Copying yields whatever the glyph slots happen to be, which for Bangla usually means nonsense.
How to tell: you can select individual words, but pasting produces junk.
The fix: PDF to Text extracts what is extractable and will show you quickly whether anything usable is there. If it isn't, treat the page as an image and run OCR — that path ignores the broken map entirely and reads the shapes.
Cause 3: it was never Unicode
The PDF was made from a Bijoy/SutonnyMJ document, so the underlying characters are ASCII letters that only look Bangla in that font. Copying gives you Avwg-style output that is technically faithful — it just isn't Bangla.
How to tell: the paste is Latin letters and punctuation, not boxes and not empty.
The fix: run the pasted text through Bijoy to Unicode. There's a fuller explanation in the legacy-font guide.
Getting a better result from OCR
If you end up on the OCR path, input quality decides everything:
- Scan at 300 DPI; higher is not better and slows things down.
- Straighten the page first — even a few degrees of skew hurts conjuncts.
- Prefer greyscale over aggressive black-and-white thresholding, which eats the thin strokes Bangla depends on.
- Crop away margins and stray marks before running it.