None of these photos were taken. Every image was generated from a few ordinary reference photos of the same person, and the point is that he still looks like himself — no smoothing, no slimming, no studio headshots. Examples by scene: travel and outdoors, food and drink, sports and fitness, coffee and cafes, beach and water, pets, nightlife, music and dance, arts and crafts, winter and snow, motorcycles, and celebrations.
Every photo in the gallery was generated for the same man from the same small set of reference photos. That is the point of showing them together: the scenes change completely while the face stays recognisably his.
No skin smoothing, no jaw slimming, no teeth whitening, no studio-headshot lighting. If a generated photo looks better than the person, it has failed at the only job it has.
Travel and outdoors, food and drink, sports and fitness, coffee and cafes, beach and water, pets, nightlife, music and dance, arts and crafts, winter and snow, motorcycles, and celebrations. The full scene catalogue in the app is considerably larger.