How to convert a Hindi PDF to text (Unicode)

Hindi documents are everywhere: court orders, government notices, land records, school certificates, newspapers, books. Getting editable text out of a Hindi PDF is often harder than it should be. Copying gives you nothing, or a string of strange English letters.

This guide explains why, and how to get proper Unicode Hindi text that you can edit, search, translate or paste into an AI assistant.

First, what kind of PDF is it?

Open the PDF and try to select a line of text.

• You can't select anything: it's a scanned PDF, a picture of the page. You need OCR.
• You can select it, and it pastes as Hindi: it already has a Unicode text layer. Any converter will extract it.
• You can select it, but it pastes as gibberish such as vkids instead of आपके: the PDF uses a legacy font such as Kruti Dev or similar. The text looks right on screen but is stored as English letters that the font draws as Hindi.

Scanned Hindi PDFs

1. Open the Docuflat converter.
2. In the OCR language picker, choose Hindi. If the document also has English (common in official forms), add English too.
3. Drop the PDF on the converter. Pages without text are recognised automatically.
4. Copy the result or download it as a Markdown file.

The conversion runs in your browser, so the document isn't uploaded. That matters for legal and personal records.

Docuflat's Hindi OCR page has more on the kinds of documents it handles well.

Legacy-font PDFs (the gibberish problem)

When a PDF uses a legacy font, extracting its text faithfully gives you the gibberish, because that's literally what's stored. There are two ways around it:

Option 1: OCR the page image. Screenshot the pages, or export them as images from your PDF viewer, and drop the images on the converter with Hindi selected. OCR reads the shapes on the page, so the font encoding doesn't matter.

Option 2: use a font converter. Tools that convert Kruti Dev to Unicode can remap the characters. They work well when you know the exact font, but mixed fonts in one document can trip them up.

For most people, option 1 is simpler and more reliable.

Getting the best accuracy

• Resolution: scan at 300 dpi if you can. Phone photos should be sharp, flat and evenly lit.
• Choose only the languages present. Adding unnecessary languages slows OCR and can cause mix-ups.
• Straighten the page. Skewed lines reduce accuracy, especially for the matras above and below letters.
• Stamps and handwriting. OCR is built for printed text. Handwritten notes and overlapping stamps will need checking by hand.

Our guide 10 ways to get more accurate OCR has more detail.

Checking the result

Hindi OCR is good on clean print but not perfect. Pay attention to:

• Similar letters such as घ/ध, ब/व and म/भ.
• Conjuncts (संयुक्ताक्षर), which can split into separate letters.
• Numbers, whether written as १२३ or 123. Check amounts and dates against the original.
• Names and survey numbers in land records, where one wrong character matters.

What you can do with the text

• Edit it in Word or Google Docs (paste the Markdown, or convert it later).
• Search long documents for names, dates or case numbers.
• Translate it, or ask ChatGPT, Claude or Gemini to summarise it. Giving the assistant text is far more reliable than a scanned PDF.

Other Indian languages work the same way. See OCR for Indian languages.