Somewhere on a modeling forum this month, someone suggested our product photos are AI-generated. We are not going to argue with a stranger in a thread. We would rather show you what to look for — because generated images of diecast models have genuinely gotten good, and the question is a fair one.
Here is the thing about AI images, though. A generator can paint you one flawless photo of a fighter jet. What it cannot do is keep the same aircraft consistent across eight different angles. The tail code drifts. A panel line appears on one view and vanishes on the next. That is not a rendering quirk you can prompt away. It is the difference between drawing an airplane and owning one.
Every listing on this site carries between seven and eleven photos, and they all come from one physical model. Our 1/72 F-14 Tomcat in VF-84 markings has ten. In every one of those shots the tail reads AJ 200, the Jolly Rogers skull sits in the same spot, and the intakes are the same shape. You can check that in a minute. A generated set cannot survive the same check — that is the whole reason we shoot instead of generate.
What follows are the seven places generated model images fall apart fastest. They work on any store's photos, including ours.

1. Letters and numbers — the three-second test
Zoom in on any letter painted on the airframe. On a real photo the edges of the digit stay crisp even when the letter is small; that is tampo printing sitting on a metal surface. Generated images mangle small text. Numbers gain extra strokes, letters lean, squadron codes melt into the paint around them. The tail of our F-4C Phantom reads NAVY in clean block capitals, with VF-83 and an AE under it — if a generated image attempted that much text at once, at least one character would come out wrong.
2. Landing gear and wheel wells
Landing gear is the most structurally complicated part of any aircraft model, and generators handle it the way someone who has never knelt next to a Tomcat would draw it. Legs appear where there is no bay. A gear-down shot shows three struts where the real aircraft has a different arrangement. Doors sit half-open at angles that would snap off the first time you touched them. Our F-14 photos show the gear both up and down, and the same nose strut and twin mains appear in every shot, in the same places.
3. Pylons and stores
Look at our 1/100 A-10. Under each wing sit separate pylons, and on the pylons hang separate bombs and missiles; you can count them and see daylight between the store and the wing surface. Generated images tend to fuse all of it into the wing, or grow a full rack of bombs on one side with nothing to balance it on the other.
4. Propeller blades
A B-17 has four engines. Each engine carries a three-bladed propeller — twelve blades in total on our 1/72 B-17G, all separate, all the same shape, all countable in the photos. Generators blur propellers into a disc, or produce six blades where three should be, or bend one blade at an angle no casting would survive.
5. Pitot tubes and antennas
The pitot tube is a needle. On a real photo it reads as a thin bright line against the background, sometimes almost invisible. Generated images make it thick, or duplicate it, or forget it entirely. Antennas get the same treatment. The small parts that would snap off in packaging are precisely the parts a generator has never held.
6. How the metal reflects
Diecast metal reflects the room it was photographed in — soft window light from one side, the grey of the background paper below. Generated metal tends to glow evenly, like painted plastic: highlights at the same intensity everywhere, with reflections that belong to no room at all. It is subtle until you have compared a few hundred of both. Then it is obvious.
7. Contact points and shadows
A model on a stand touches the stand at one small point, and that point carries weight. Where the fuselage meets the desk, the shadow is tight and dark and matches the light direction of everything else in the frame. Generated images float. The shadow falls one way, the highlight falls another, and the patch where the stand meets the belly smears into vague grey.
Scale matters for how many of these details survive at all — we put the 1/72 and 1/100 sizes side by side in an earlier guide.
Why we photograph instead
Honest answer: photographs are slower, and they show more flaws. Our photos carry mold seam lines, the faint edge of a decal, the occasional fingerprint of dust that survived the wipe-down. They are not glamorous. One of our reviewers put it better than we could: "Finish is better than the photos in description. Details sharp and distinct." — E. Jones, on the Super Hornet listing. He liked the model more than our photography of it. We would rather fix the photography than replace it with something that never existed.
The compromise we made is this: every photo on a listing is of the same physical model, shot the same day, on the same paper. Not a render, not a supplier image, not a generated scene. If two shots disagree about where a panel line runs, that is our problem to fix — and it is at least a real disagreement about a real object.
One last thing worth knowing, and it works on any store including ours: run a product photo through a reverse image search. Photos shot in-house tend to surface once, on the store that took them. Supplier renders and generated scenes surface on dozens of listings at once, often with the identical background and lighting. It takes about ten seconds and it tells you more than any product description will.