People have had tons of fun finding collisions and seeing how far they can take an image until the apple neuralhash algo thinks it different. It very much is no

People have had tons of fun finding collisions and seeing how far they can take an image until the apple neuralhash algo thinks it different. It very much is not an exact match like what you would get with MD5 for example. But a "perceptual hash" that means that it gets the same has even if you crop it by a couple of pixels or change some pixels.