Image similarity · image
How similar are these two images?
Two files can be byte-for-byte different, a different size, a different format, a fresh re-save, and still be the same photograph. This compares them by what they look like, not their bytes, using a perceptual hash, and tells you how close they are.
Runs entirely in your browser. Your images are never uploaded or stored.
What a perceptual hash is
A normal hash like SHA-256 changes completely if a single byte changes, so it can only tell you two files are identical. A perceptual hash is the opposite: it shrinks the image to a tiny grayscale grid and records the pattern of light and dark, so two pictures that look the same produce almost the same hash even after resizing, re-compression or a format change. Comparing the two hashes bit by bit gives a distance: zero means the fingerprints match, a few bits means a near copy, many bits means different images.
What the distance tells you
A distance of zero means the two look identical to the hash, typically the same image re-saved or converted. A small distance usually means the same picture that has been resized, recompressed or lightly adjusted. A large distance means the images are different, or one has been cropped, heavily edited or is a genuinely separate shot. Because the hash works on a coarse grid, it is strong at catching re-saves and rescales and weak at crops, so read a large distance as "not obviously the same" rather than proof they are unrelated.
Uses and limits
This is the check for "is this the same image that was published elsewhere," or "is this so-called original really just a re-export of that one." It does not tell you which came first, and it is not a reverse image search. To dig into a single file instead, see the photo edit check and the metadata viewer.