I build one of these. The useful version of that is telling you where it does not apply, because the marketing in this category will not.
Where a person in the room wins
Hard faces. These models reproduce well what the base model saw a lot of and average what it saw little of. Tightly curled or high-volume hair, deeper skin tones, strong prescription glasses, pronounced asymmetry.
A photographer sees that a particular face needs the key light moved and moves it. There is no software equivalent, and prompt engineering is not one. This is the single clearest boundary.
When the brief is a reportage rather than a portrait. Space, work, detail, product. That is a session, not a headshot, and booking the wrong one is the most expensive common mistake here.
When one image has to be right. A book jacket, a keynote bio, anything where you get one shot at it. Volume is the generated route's advantage, and volume does not help when the requirement is a single specific frame.
Where the generated route wins, honestly
Volume, consistency and reproducibility eighteen months later for someone who has not joined yet. That is a narrower claim than "better photographs" and it is the true one.
The budget comparison against studio work sets out the numbers, and the broader comparison with traditional photography covers the rest. Both are ours, so weigh accordingly.
The effect size question
The claims made for professional photographs are usually too strong. A portrait does not make anyone more qualified and no one hires on one.
What it does is stop someone not clicking, which is a smaller and more defensible claim. Most bad profile photographs are not unflattering, they are ambiguous: cropped so the head is cut, lit so the eyes fall into shadow, framed so wide the face is a smudge at the size it is actually rendered.
What a professional photograph does for a profile is the honest version, and it is a smaller effect than the category usually implies.
Two practical resources
If you are evaluating options, the portrait generator comparison covers the field including us.
And the questions that come up most often, with answers rather than marketing: the FAQ.
The test I would run before deciding anything
Take the hardest face you have access to, ideally your own. Fifteen photographs across four different days. Run it through a free tier.
If the result does not look like the person, nothing in this category fixes that today, and you have saved yourself the research and the money.
More on this subject, grouped by topic: Everything published here, by subject
The tool I sell, for completeness: BetterPic.
Top comments (0)