QR & Barcode Scanner

Which codes can it read?

QR (including Micro QR and rMQR), Data Matrix, Aztec, PDF417 and MaxiCode among the 2D formats; EAN-13, EAN-8, UPC-A, UPC-E, Code 39, Code 93, Code 128, Codabar, ITF, DX Film Edge and the GS1 DataBar family among the 1D ones. That is the full read set of ZXing C++, which is the engine here.

It scans in any orientation and tries the inverted image as well, so a white-on-black code — common on packaging and on screens in dark mode — reads without you having to do anything.

It found nothing in my photo. Why?

Usually one of four things. The code is a small part of a large photo, so after the image is scaled down for speed there are not enough pixels left across each bar; crop closer to the code and try again — the crop tool is next door. The photo is blurred, or shot at a steep angle so the bars merge. There is no quiet zone, because the code was cropped flush to its edge or is printed over artwork. Or it is a symbology this engine does not read — a few exist, mostly postal codes.

Large images are scaled to 2000 pixels on the long edge before scanning. That is a deliberate trade: at full resolution a modern phone photo takes tens of seconds, which people reasonably read as "broken". If a code fails and you suspect the downscale, crop first — a cropped image below that size is passed through untouched.

What happens if the code is a link?

You are shown the address as text and given a link you have to click. Nothing opens by itself, and nothing but an http:// or https:// address is offered as a link at all — a javascript: or data: payload is displayed as plain text and stays inert.

That is not caution for its own sake. A printed QR code is untrusted data from a stranger, and "quishing" — stickers pasted over legitimate codes on parking meters, restaurant tables and delivery lockers — is a real and cheap attack. Reading the address before you follow it is the entire defence, so this tool makes you read it.

Why does a Wi-Fi code show up as network and password fields?

Because the raw text of a Wi-Fi QR code is a structured record, not a message: WIFI:T:WPA;S:Cafe Guest;P:flat\;white;;. Handing you that string would be a worse answer than the two things you actually want. The parser handles the escaping rules properly, so a password containing a semicolon, colon, comma or backslash comes back exactly as it was — a naive split on ; mangles it, and passwords like that are perfectly ordinary.

vCard contact codes are broken out the same way, into name, organisation, phone, email and web address. Both formats are the ones our own QR code generator writes, so a code made there reads back into the same fields it was built from.

Can it read several codes in one picture?

Yes, up to 32 in a single image, and each one is listed separately with its symbology, its text, its position in the image and its own copy button. A shipping label with a Code 128 and a Data Matrix side by side, or a page of printed labels photographed together, comes back as a list rather than as whichever code the detector happened to find first.

What does the camera button do, and does it stay on?

Pressing it asks the browser for camera access; nothing before that point touches the camera. Frames are grabbed roughly seven times a second and decoded in this tab. The moment a code is read, the camera stops and its tracks are released — which is what turns the indicator light off — and it also stops if you press the button again or navigate away.

Browsers only allow camera access on a secure origin, so this works on https:// and on localhost, and is blocked on a plain http:// address. If access is denied, drop a photo instead; the result is identical.

Is the image uploaded anywhere?

No. The image is decoded by this page, and both the decoding and the barcode engine run in this tab. The engine is zxing-wasm, a WebAssembly build of ZXing C++; its 1.5 MB binary is served from this site's own domain rather than a CDN, and it is only fetched the first time you actually scan something. Camera frames never leave the tab either — they are drawn to a canvas in memory and handed straight to the decoder.