Scan text with your camera
Point your camera at printed text and copy it as words.
Your camera stays on this device. No photo is ever uploaded.
From a page to editable words
Point the camera at printed text, take a photo, and get the words back as text you can copy. A textbook paragraph you want to quote, a printed form you need to fill in digitally, a business card, a serial number on the back of a router.
The capture is processed first
Before recognition, the photo goes through the same document flattening the scanner uses: the lighting across the page is estimated and divided out, leaving even, high-contrast black on white. This matters more for recognition than for looks — a recogniser handed a shadowed photo loses whole lines in the dark half, and the flattened version usually reads cleanly.
Where the photo does go
Worth being straight about, because the other camera tools here never upload anything. Recognition runs on our server, so the captured frame is sent — it is processed in a directory created for that one request and deleted as soon as the text comes back. Nothing is logged and no copy is kept.
The capture and the flattening happen on your device; only the finished, processed image travels.
Getting it right first time
- Fill the frame with the text. Resolution per character is what decides accuracy.
- Shoot straight on — text photographed at an angle leans and stretches.
- Keep still. Motion blur defeats recognition more often than poor light.
- For a single paragraph, get close to that paragraph rather than photographing the whole page.
Check the numbers
Recognition fails plausibly rather than obviously — it returns a word that looks similar, not gibberish. The confidence figure is reported, and low confidence is flagged, but always check digits and names against the page: 0 and O, 1 and l, 5 and S are the usual confusions.
Whole documents
For several pages, scan them to a PDF first and then OCR the PDF — you get a searchable document rather than loose text.
Common questions
Is my photo uploaded?
Yes, for this one — recognition runs on our server. The capture and flattening happen on your device, and the image is deleted as soon as the text returns.
Why is the photo processed before reading?
Flattening the lighting removes shadows that would otherwise cause whole lines to be missed.
Does it read handwriting?
Poorly. The models are trained on printed text.
How do I improve accuracy?
Fill the frame with the text, shoot straight on, and hold still. Resolution per character matters most.
Which languages work?
Latin scripts and Chinese are handled well by the bundled models. Other scripts vary.
Can I scan several pages?
Scan them to a PDF first, then OCR the PDF — you get a searchable document rather than loose text.