Image tools AI like Midjourney, Dall · E and stable distribution are formed in millions of real photos. They learn the models how the lighting works, how faces are shaped, what the backgrounds generally look like and then use this knowledge to build new images from zero. The results often seem just at first glance.
The human cortex exists for the fastest recognition of a face, an object, a model. It works very well most of the time, but it also means that the brain often provides additional information in the event of a detail. If it seems real enough, the brain is ready to suspend the disbelief without asking too many questions. This is what makes the images IA so credible, even when they are riddled with small errors.
