What happens when you press Read
The page loads Tesseract’s LSTM engine with its Hindi and English language data from this site, once, and reads the picture in a background worker. Both languages run together, because Hindi notices, forms and forwards are rarely only Hindi. What comes back is lines of Unicode text, each with a box on the picture and a confidence.
The transcript is shown in Noto Sans Devanagari, served by this site, so it reads the same on every device. Words the reader was least sure of are outlined on the picture and listed under the text; matras and conjuncts in small type are where to look first.
Unicode, not Kruti Dev
Much Hindi typing in offices still uses Kruti Dev and similar fonts, which store Hindi as Latin letters drawn with Devanagari shapes. Text in that encoding looks right only in that font and turns into strings like “fgUnh” anywhere else. This page writes Unicode Devanagari — the same characters a phone keyboard types — so the result searches, sorts and pastes correctly everywhere.
What it will not do
- It does not read handwriting.
- It does not convert to or from Kruti Dev.
- It reads Hindi and English only; Marathi, Nepali and Sanskrit share the script but are not loaded or claimed.
Questions people ask
Is my picture uploaded?
No. The Hindi reader (Tesseract with its Hindi and English data, about 8.4 MB the first time) is downloaded from this site and runs in your browser tab. The picture never leaves it.
Will the text work in Word?
Yes. The text is Unicode Devanagari, the encoding Word, Google Docs, WhatsApp and every web page use — not a legacy font encoding like Kruti Dev, which turns into gibberish outside its own font. “Copy for Word” puts it on the clipboard with a Devanagari font named, and the .txt download carries a byte-order mark so Notepad and Word open it as Unicode.
Can it read Hindi and English on the same line?
Yes. Both languages are loaded together, so a circular with an English heading, a form with English field names or a forward that mixes the two is read line by line in both.
How accurate is Hindi OCR?
On clean print — a typed circular, a newspaper column, a notice board photographed straight on — it gets most words right. Conjuncts and matras in small or decorative type, low light and photographs at an angle cause most errors. Every word carries the reader's own confidence and the uncertain ones are outlined so you can check them. The site has not yet published a measured figure, so it does not quote one.
Does it read handwritten Hindi?
No. The reader is built for print. On handwriting it reports low confidence across the page rather than presenting a guess as text.
Where to take this next
- Image to textRead the printed English or Hindi text in a picture, in this tab, and copy or download it.
- Screenshot to textRead the text in a screenshot, and lay a chat out as Them: and You: lines.
- Images of textWhen a picture's content is its words, the alt text is the words. WCAG 1.4.5 and the limits of OCR.
- WCAG 1.1.1 and GIGW, in plain termsWhat the success criterion asks for and what an Indian government audit checks.
