The one prediction that looks safe

Forecasts about generated photography are mostly wrong in the same direction: they predict replacement and describe a battle that is not happening.

What has actually settled

Image quality plateaued. Almost every product produces realistic headshots, which means realism stopped being a differentiator and the comparison articles all started reading the same.

Where the competition moved

Consistency across a group of people over time. Producing one good portrait is routine. Producing the forty-first, fourteen months later, so that it matches the first — that is unsolved and duller and where the actual work is.

What did not change

Where the failures cluster: strong prescription glasses, tightly curled hair, darker skin tones, pronounced asymmetry. Two years of progress has not moved those, across any vendor.

The thing photographers still do that nothing reproduces

Watch you and say your chin is slightly too high. For one important headshot that is worth the whole fee, and no forecast changes it.