the one thing I know about AI images is that they are very anatomically-consistent when generating crowds of people. they do not decouple into flesh beasts. they are all very normal, extremely physical, non-duplicative, and are not the first thing you would look towards to see if an image of a zoomed-out crowd of people was AI-generated.

In fact, it's perfectly intuitive that a lossy parrot would be able to consistently draw the photorealistic anatomy of a human 100+ times in the same photo at a minuscule scale and would not, in fact, have any noticeable blips at a scale of single-digit pixels. Or, given downscaling, low double-digits at most.

It would also perfectly generate and perspective-warp the text to the bent signs with a campaign logo that was unveiled less than a week ago, which is enough time for the best and most-consistent models to already have integrated it into their dataset.

It is also very consistent with regards to physicality and would

generate

Show

the

Show

background

Show

consistently

Show

at

Show

different

Show

angles

Show

and of course, the one thing I know to trust my posting peers on the internet to know is how mirrors work. 100% track record. 100% consistency. It's the one thing everyone truly understands. Entirely intuitive. Something they don't even offer college courses on because even elementary schoolers fully understand exactly how reflection works. Yep. mhm.

Show