Turn printed Hindi in photos, scans and screenshots into text you can copy and edit, together with the English words and numbers that Indian documents mix in.
No limits or paywall.
Use instantly, no account.
Processed inside your browser.
Clean, full-quality output.
No upload queue or waiting.
Works on phone and desktop.
Retyping Hindi from a photo is slow, especially with a Hindi keyboard you rarely use. Notices, circulars, forms, textbooks, receipts and government letters are full of printed Hindi that you may need to copy, translate, search or edit. OCR (optical character recognition) reads the text for you.
This free tool recognises printed Hindi in Devanagari script together with the English words and numbers Indian documents mix in. Drop or paste an image and get editable text in a couple of seconds. Recognition runs entirely in your browser, so photos of personal documents are never uploaded.
Before
Photo of a printed page
Image
Four lines of Hindi with a date, a PAN number and a phone number, slightly tilted and blurred.
After
Editable text
≈ 95% exact
Every Hindi word, the date, PAN and phone number read correctly in about 2 seconds.
Our test: a clean screenshot was read with 0% character errors and a simulated phone photo with 4.8%, using the Hindi + English setting. Accuracy depends on the image; handwriting is not supported.
Drop a JPG, PNG, WebP, BMP or TIFF image into the tool above, or paste a screenshot with Ctrl+V.
The Hindi + English setting is preselected, so Hindi text, English words and numbers are all recognised.
For grey or faded scans, turn on grayscale or the threshold filter before extracting.
Press Extract Text Now, check the result, then copy it or download it as a .txt file.
| Language setting | Clean screenshot | Phone photo (simulated) | Numbers and PAN |
|---|---|---|---|
| Hindi + English (default here) | 0.0% | 4.8% | All correct |
| Hindi only | 6.2% | 26.7% | PAN and digits misread |
| English only | Hindi becomes gibberish | – | Correct |
The tool uses Tesseract, the open-source OCR engine originally developed at HP and Google, compiled to run in the browser. It finds lines of text in the image, splits them into words and recognises the characters with a neural network trained on printed Devanagari. Hindi is harder than English for OCR: letters join along the shirorekha (the headline), and vowel signs (matras) and conjuncts sit above, below or beside consonants. Modern models handle this well for printed text.
Indian documents rarely contain Hindi alone: names, PAN and Aadhaar numbers, dates and phone numbers usually appear in Latin script and Western digits. A Hindi-only model reads those badly. That is why this page preselects Hindi + English, with English first, which in our tests read the Hindi words and every number correctly.
We tested with four lines of Hindi containing a date, a PAN number and a mobile number, rendered as a clean screenshot and as a simulated phone photo (tilted 1.2°, blurred, grey paper, heavy JPEG compression). With Hindi + English, the screenshot came out with no character errors and the photo with 4.8%, almost all of it stray marks from the edge of the page. All Hindi words, the date, the PAN and the phone number were exact in both.
OCR accuracy depends mostly on the image. A few minutes of preparation pays off:
People use Hindi OCR on certificates, ration cards, letters and forms, which are personal documents. Here, the image is processed by your own browser. The only download is the language data itself (about 1.4 MB for Hindi and 10.9 MB for English, compressed), fetched once from the jsDelivr CDN and then cached. Your image is never sent anywhere.
Marathi, Nepali, Konkani and Sanskrit share the Devanagari script, so much of their text is recognised with the Hindi model. Characters and spellings specific to those languages are more likely to be misread, so proofread the result. For other scripts and languages, the main Image to Text tool also offers Spanish, French, German, Chinese, Arabic, Russian, Japanese and more.
The output is standard Unicode Devanagari, so it works everywhere modern text does: paste it into Word or Google Docs, search it, send it in WhatsApp or email, or put it into a translation tool to read it in English. Because it is real text, it can also be read aloud by screen readers, which helps people who find small printed Hindi hard to read.
OCR output keeps the line breaks of the image. Before reusing a long passage, join lines that were broken only because of the page width, and proofread numbers and names carefully.
OCR works on images, one page at a time. For a scanned Hindi PDF, convert the pages to PNG with PDF to PNG at 300 DPI, which gives the OCR engine enough detail for small matras and conjuncts, then run each page through this tool and paste the results together. For a long document, process the pages you actually need first; each page takes a few seconds once the language data is cached.
Copy text from circulars and notices to search, translate or reply.
Digitise Hindi textbook pages and question papers.
Extract names, dates and numbers from printed Hindi documents.
Copy Hindi quotes or messages from screenshots with Ctrl+V.
Digitise printed Hindi documents such as old letters, wedding cards, temple or society notices so they can be searched and shared as text.
Drop the image into the tool above (or paste a screenshot with Ctrl+V), keep Hindi + English selected and press Extract Text Now. Copy the text or download it as a .txt file.
Most Indian documents mix scripts: Hindi text with English names, PAN and Aadhaar numbers, dates and phone numbers. In our test, Hindi-only mode turned the PAN number into Devanagari-looking characters and dropped digits, while Hindi + English read everything correctly.
On a clean screenshot of printed Hindi, our test text came out with no errors. A tilted, slightly blurred photo of the same text had a 4.8% character error rate, mostly stray marks near the edges. Always check names and numbers before using the text.
Not reliably. The OCR engine (Tesseract) is trained on printed text. Handwriting usually produces poor results.
No. Recognition runs in your browser. The first time, your browser downloads the Hindi and English language data (about 12 MB) from a public CDN and caches it; your image never leaves your device.
They use the same Devanagari script, so many words are recognised, but the language model is Hindi, so unusual characters and spellings can be misread. Check the result carefully.
Yes. Copy a screenshot (for example with Windows + Shift + S) and press Ctrl+V on the page; it loads straight into the tool.
The first run downloads the language data. After that it is cached, and an image like our test page takes about 2 seconds.
Convert the PDF pages to images first with PDF to PNG, then run OCR on each page image.