A PDF is a picture of a page layout, not a document you can edit. That is true for English and it is especially true for Hindi written in a legacy font such as Kruti Dev. Sometimes the text comes out as Hkkjr; sometimes it comes out as boxes; sometimes it does not come out at all. The cause is almost always the kind of PDF you have, not the file being damaged.
This guide walks through the three kinds of PDF you are likely to meet, what to do with each, and how to finish with clean Unicode text in Word.
First, work out what kind of PDF it is
There are three common cases.
1. A PDF with a real text layer in a legacy font. The PDF was exported from Word or a DTP program where the text was typed in Kruti Dev. The letters are stored as English characters, but the PDF records which font was used. A good extractor can see the font name and run the text through the matching converter.
2. A PDF with a real text layer in Unicode. The PDF was made from Unicode text. Extraction gives you real Devanagari, and there is nothing to convert. You may still have to fix line breaks and spacing.
3. A scanned PDF or a photo of a page. There is no text layer at all, only pixels. No extractor can read letters that were never stored as letters. You need OCR (optical character recognition).
You can usually tell case 3 by trying to select text in a PDF reader. If the selection is empty or selects the whole page as one image, it is a scan.
Case 1 and 2: use the PDF to Unicode converter
Open the PDF to Unicode converter. Choose the PDF (the page accepts files up to 40 MB), and let it read the text layer.
The tool does three things:
- It reads the text and font names page by page.
- It recognises legacy font runs and decodes them. Font names in PDFs often carry a six-letter subset prefix such as
ABCDEF+, so the tool strips that before matching the font. - It keeps runs that are already Unicode as they are.
It does not guess when the font name is generic or missing. That is deliberate: a wrong guess produces confident-looking but incorrect Hindi.
When the extraction finishes, review the page report before you download anything. Then copy the editable text, or download it as .txt or .docx.
The tool extracts text. It does not produce an edited PDF, so do not expect the layout to survive. What you get is paragraphs, line by line, which you can then rebuild in Word.
Case 3: scanned pages need OCR
If the PDF has no text layer, use the image to text tool. It can open a PDF, render each page as an image, and recognise the text with a Hindi model. You can also upload a photo or paste an image.
A few practical points:
- Recognition is built for printed text. Handwriting, ornate fonts and low-resolution scans need a lot of correction.
- Keep the page straight and well lit if you are photographing it.
- The confidence score is the engine’s estimate, not a guarantee. Proofread names, numbers, vowel signs and conjuncts.
- Choose the right language model. Hindi text needs the Hindi model, not English.
OCR of Devanagari is useful but not perfect. Treat its output as a first draft that you correct, especially around the ि sign and conjuncts such as क्ष.
Clean up the converted text
Whether the text came from a text layer or from OCR, check it before you rely on it. Common problems in extracted PDF text include:
- Columns and tables in the wrong order. PDF stores text in drawing order, which may not match reading order.
- Headers and footers mixed into paragraphs.
- Words split across lines, with hyphens or without.
- Spacing lost around punctuation.
- Mixed pages: some pages with a text layer, some scanned.
If a page looks wrong, convert that page separately rather than fighting the whole file.
If you only have a paragraph, not a PDF
If you have already copied the text out and it looks like Hkkjr, skip the PDF step and paste it straight into Kruti Dev to Unicode. That converter is the right tool for a block of legacy text from any source: a PDF, an email, a website or an old file.
If you are not sure which legacy font produced the text, paste a sample into the font detector first. Kruti Dev and DevLys share most key positions, so very short samples may fit both; a sentence or two gives a better answer.
If the whole document is a Word file
If you have the original .docx rather than a PDF, do not extract it at all. Use the Word file converter. It reads the font of each run, converts legacy text to Unicode, and keeps tables, images and styling. It also gives you a report of what it changed, which is far less work than rebuilding the document from extracted paragraphs.
Finish in Word
Once the text is Unicode:
- Paste it into Word.
- Select it and set a Unicode Devanagari font such as Mangal, Nirmala UI or Noto Sans Devanagari.
- Rebuild headings, lists and tables.
- Run a spell-check if you have one for Hindi. Legacy text could never be checked; Unicode can.
Do not simply change the font of the extracted text to Mangal and hope for the best. Changing the font changes how the letters are drawn, not what they are. If the underlying characters are still the old English-letter codes, you will see Hkkjr in a different typeface.
A short checklist
- Can you select the text in the PDF? If yes, it has a text layer.
- Does the extracted text look like
Hkkjr? It is legacy; convert it. - Does it look like real Hindi? It is already Unicode; just clean it up.
- Is the page a scan? Use OCR, then proofread.
- Is the source a Word file? Convert the file, not the PDF.
Working in that order saves the most time and avoids the most common mistake: running OCR on a PDF that already had perfectly good text inside it.