Before You Post - Photo Privacy Check

No. The metadata check, face detection and text reading all run in your browser. The first time, your browser downloads the face detector (about 4.6 MB, from the jsDelivr CDN and Google's model storage) and the English text-reading data (about 3 MB, from jsDelivr), then caches them. Those downloads contain no part of your photo.

EXIF is the metadata standard cameras and phones use to store details inside a photo file. Its GPS section (CIPA DC-008, the Exif 2.32 standard) has tags for latitude, longitude, altitude and even a horizontal positioning error in metres. When location is allowed for the camera app, the phone fills these in, and anyone who has the original file can read them with free software.

Blurred or pixelated text can sometimes be read back: a 2016 study by McPherson, Shokri and Shmatikov showed neural networks recovering content from pixelated and blurred images. A solid black box leaves nothing to recover. Faces get Blur Faces' effect, which shrinks each face to a handful of pixels before blurring, so the detail is gone rather than softened.

No, and nothing automatic can. It reads metadata, finds faces and reads printed English text, then looks for known shapes: card numbers, bank account numbers, emails, phone numbers, street addresses, API keys and long random-looking codes. It can miss text that is small, blurry, rotated, handwritten or in an unusual font, and it can't judge context: reflections in glasses or windows, a house number or street sign, a uniform, a badge, a landmark or a view from your window. Look over the cleaned image yourself before you post it.

It checks the Luhn checksum that payment card numbers use (ISO/IEC 7812), and bank account numbers (IBAN) against the mod-97 check from ISO 13616. A passing checksum means the number is well formed, not that the account exists - test numbers like 4242 4242 4242 4242 pass too.

All metadata: EXIF (including GPS, device, serial numbers and timestamps), XMP, IPTC captions and the hidden preview image. The image is redrawn on a canvas and saved again, and canvas output carries no metadata. After saving, the page reads the new file back and tells you how many metadata fields it still has. Ticked faces are blurred and ticked text is covered by black boxes.

Dimensions stay the same. JPEG and WebP files are saved again at 92% quality, which is very close to the original but not bit-identical; PNG stays lossless. The original colour profile isn't kept (browsers save a standard sRGB one or none), so photos shot in a wide colour space can look very slightly less saturated.

Yes. Safari opens them directly; other browsers convert them on your device first with a WebAssembly build of libheif (about 1.3 MB, loaded only when needed). Browsers can't save HEIC, so the cleaned file is a JPEG.

Some sites remove metadata when you upload, but you can't see what each one keeps, and a photo sent as a file attachment, through a cloud link or by some chat apps' send-as-file option is passed on as it is. Cleaning it yourself first means it doesn't matter where it goes.