with face scanning, you’re getting a digital replica of a real person’s head. nothing gets more realistic than that. then the artist optimizes it, removes glitches, refines details like pores, wrinkles. when all things are ready, they retopologize it for a lower resolution game-ready mesh and bake the sub maps(ambient occlusion, normal, displacement, etc) from the high polygon model.
this is advantages because the process is faster than creating a head of the same quality and realism from scratch. you get more heads, from which adjustments can be made to both texture and sub maps, and to the model itself, creating many permutations in less time. very important for open world with many npcs.
complete procedural technology lacks both the pure representation of face scanning, and lacks the artist’s detail and touches. it simply cannot approach any level of life-like appearance, and ultimately can only generate barby doll like faces with some organic features, but overall, not very human looking by comparison to what artists can do, or better yet what artists can do with aid of face scanning.