If you research portrait generators you will find a dozen roundups that disagree about which is best. It looks like corruption and part of it is: this category runs on affiliate commission to an unusual degree.
We run a programme too, so this is not a stone I am throwing from any height. Our affiliate terms are public precisely so the conflict is visible rather than implied.
What follows is how I would read any of it, ours included.
Separate the verdict from the method
The verdict is the part that commission influences. The method is usually stated honestly because it is boring.
If a comparison does not tell you whose faces, how many source photographs, and when it ran, the conclusion does not transfer to you. Those three variables move results more than the difference between products.
The disagreement that survives honesty
Even with no commission involved, four choices produce different winners.
Which faces. These models fail unevenly. A test on three easy faces measures a tie.
What counts as good. Photographic quality and likeness come apart, and products trade one for the other.
Input preparation. Fifteen photographs from one afternoon degrades every product, so a review doing that ranks tools by graceful degradation instead.
When it ran. Queue depth varies with load, so a turnaround figure is a measurement of a moment.
What we publish and how to weight it
The 2025 state of the product and the general guide are ours. The work-context version covers what a professional portrait has to do at the office rather than on a profile.
The craft writing is where a participant is least conflicted, because lighting and framing do not vary by vendor: the photography insights collection is the part I would actually recommend.
The test that outranks all of it
One hard face, ideally yours. Fifteen photographs across four different days. Two products the same afternoon. Then a blind comparison: mix five generated with five real and ask someone who knows that face to separate them.
Twenty minutes, no commission involved, and it samples the only part of the distribution that matters to you.
What I would want from this category
Published keeper rates. Usable outputs over total outputs, per product, on a stated test set. It is the number that decides satisfaction and nobody publishes it, us included, which is a fair thing to hold against all of us.
Disclosure, in the spirit of the piece: I work on BetterPic.
Top comments (0)