Hindi PDF to Word and Excel

What works

Printed Hindi text on scanned pages is read with OCR, page by page. If your page also has English words, choose the Hindi + English option in Convert.
For digital PDFs (with real text, not a photo), Convert puts each table into its own Excel sheet. The sample below is a scan, so it does not test this.
You get both a Word file and an Excel file from one upload.
Files up to 15 MB and 40 pages each.

Tables on scanned pages are not rebuilt: Convert writes each line as a plain Word paragraph, so table text comes out as loose lines, not cells.
Very faint or handwritten scans may come out rough.
English words on a page converted with Hindi only can come out garbled. Use the Hindi + English option for mixed pages.
Check important results against the original.

Real output, 2026-10-15. One test page with a heading, two paragraphs and a small table, put through Convert as a scan. Nothing below is edited.

Text accuracy on this page: 87.9% of characters match the source (measured as 1 - Levenshtein/len, whitespace removed, vs source strings). One page is not a benchmark.
Table kept as a table: no.
Body paragraphs read well (minor nukta/diacritic slips); Latin heading garbled by Hindi-only model.
Table header 'क्रम संख्या' misread.
Convert's OCR path writes every line as a plain Word paragraph: the 3x3 table was NOT rebuilt as a Word table (docx has 0 tables); cell text survives as loose lines.
Page-level OCR only; xlsx from Convert was fetched but not evaluated.
after.png is LibreOffice 'soffice' render of a render-only copy of sample.docx with the default font swapped to Noto (LibreOffice here had no Indic fallback font); sample.docx itself is unmodified.

2 free pagesThe first 2 pages of every file, no sign-up.

₹100 per pageEach extra page, paid by UPI in India.

£1.00 per pageEach extra page, paid by card outside India.

Deleted on a timerFree files after 24 hours, paid orders after 7 days.

Never used to train AIYour files are processed, returned and deleted.