Hindi, Gujarati or Tamil PDF converts to gibberish: the fix
You open a Hindi circular, a Gujarati land record or a Tamil newsletter. It looks perfect on screen. But when you copy a line, or convert the PDF, you get something like Hkkjr ljdkj instead of भारत सरकार.
Nothing is wrong with the converter. The PDF really does contain those Latin letters. This guide explains why, how to recognise these files, and the simplest way to get proper text out of them.
Fonts that pretend to be a script
Before Unicode became the standard way to store Indian scripts (roughly the late 2000s, and later in many offices), people typed Indian languages with special fonts. Each font took the ordinary keys of an English keyboard and drew Indian letters in their place. Type H, k, k, j, r in Kruti Dev and you see भारत.
On paper and on screen this works. But the file only stores the keys that were pressed. Anyone without that exact font, and any software that reads the text itself (search engines, AI tools, converters, screen readers), sees the Latin letters.
These fonts were everywhere, and documents typed in them are still being created and shared as PDFs:
| Language | Common pre-Unicode fonts and tools |
|---|---|
| Hindi | Kruti Dev, DevLys, Chanakya |
| Marathi | Shree-Lipi, DV-TT fonts |
| Gujarati | Terafont (Varun, Akash), LMG Arun |
| Bengali | Bijoy with SutonnyMJ and similar fonts |
| Tamil | Bamini, TAB and TAM fonts, TSCII |
| Telugu | Anu fonts |
| Kannada | Nudi and Baraha (older versions) |
| Malayalam | ML-TT fonts |
| Punjabi | AnmolLipi, Asees, Joy |
| Nepali | Preeti, Kantipur |
| Urdu | InPage |
How to tell what you have
Open the PDF and try to select and copy a line, then paste it into a notes app or an email.
- It pastes as the right script. The PDF uses Unicode. Any converter, including Docuflat, will extract it directly and perfectly.
- It pastes as Latin letters, numbers and symbols that don't make sense. It's a legacy-font PDF.
- Nothing can be selected. It's a scan: a picture of the page. You need OCR anyway; see converting a scanned PDF.
Some documents mix both: a Unicode letterhead with a body typed in Kruti Dev, or English in Times New Roman with a Gujarati paragraph in Terafont. Check a line from each part.
The reliable fix: convert pictures of the pages
OCR doesn't care how a file stores its text. It looks at the shapes on the page, exactly as you do, and writes out real Unicode letters. So the simplest fix for a legacy-font PDF is to make the converter read it as a picture.
- Turn the pages into images. Take a screenshot of each page at a comfortable zoom level. Or export the pages as images: on a Mac, Preview's File → Export saves a page as PNG, and many PDF viewers on Windows have an "Export to image" option. (Printing to a new PDF doesn't help: it copies the same broken text.)
- Choose the language in the OCR box, plus English if the page mixes both.
- Drop the images onto the converter. Several at once is fine.
- Copy or download the text. It's now real Unicode, so it can be searched, edited, translated and pasted into AI tools.
Screenshots work well because the text on screen is sharp. Zoom the PDF so that body text is comfortably large, around the size you'd read it at, before capturing.
The other fix: font converters
There are free tools that convert Kruti Dev to Unicode, Bamini to Unicode, Preeti to Unicode and so on. They map each stored Latin character back to the letter the font drew. They are fast and exact when:
- you know exactly which font was used, and
- the whole document uses that one font, and
- you can copy the text out of the PDF in the first place.
They go wrong when a document mixes fonts, when the PDF's text order is scrambled (common in multi-column layouts), or when the font is a variant the converter doesn't know. Then you get a mixture of correct words and new gibberish, which is harder to spot than the original problem.
For most people, OCR on page images is simpler and more predictable. Font converters are worth it for very long, single-font documents where you need every character exactly as typed.
Checking the result
OCR on clean, screen-sharp text is accurate, but check the details that matter: names, amounts, dates, survey and case numbers. Our proofreading checklist shows a quick way to do it. Indian-language digits (such as १२३ or ૧૨૩) are kept as printed; convert them to 0–9 with find and replace if you need them in a spreadsheet.
If you create documents
If you or your office still type in a legacy font, switching to Unicode saves everyone downstream this trouble. Windows, macOS, Android and iOS all include Unicode keyboards for Indian languages (for example the Hindi Traditional or INSCRIPT layouts on Windows), and phonetic keyboards such as Gboard let you type Indian languages by spelling words in English letters. Documents typed this way can be searched, read by screen readers and understood by AI tools without any conversion at all.