Read the text off scanned pages, in twelve Indian languages and several others, with pages rendered at print resolution first for accuracy.
Runs in your browser — nothing is uploaded
Drop a scanned PDF here
Nothing is uploaded — the text is read on your device.
Your file never leaves this device. The whole operation runs in your browser — nothing is uploaded, queued on a server, or deleted later, because nothing was ever sent.
Recognised text
—
Load a scanned PDF to begin
Rendered at 300 DPI before reading
Twelve Indian languages, and the model comes to you
OCR accuracy is dominated by input resolution — this is why other browser OCR tools disappoint.
Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi and Urdu.
You are told how sure the recognition was, rather than handed text of unknown quality.
The conversion happens in your browser. Nothing is uploaded, so nothing has to be trusted or deleted later.
Drop in the scanned PDF.
Choose the language of the document — this matters a great deal.
Read a few pages at a time, then copy or download the text.
Optical character recognition works by matching shapes. At screen resolution — 96 DPI or below — the strokes that distinguish a 5 from an S, or a त from a न, are only a pixel or two wide, and the match becomes a guess.
Pages here are rendered at 300 DPI before recognition. It costs a few seconds a page and it is the difference between usable text and noise. Tools that skip it are the reason people conclude browser OCR does not work.
Each language has its own trained model. Running an English model over Devanagari produces confident nonsense rather than an error, because the recogniser will always find *some* Latin letter that resembles each shape.
If the output looks like gibberish, check the language before anything else.
The recognition model — roughly 15 MB for a language — is fetched once and then cached by your browser. That download is the only network traffic involved.
Your document is never sent anywhere. The model comes to your document, not the other way round, which is the opposite of how every OCR service works.
Available tools are linked; the rest are in development.
We build internal tools, dashboards and automation that fit how your team works.