The method is deliberately boring, because a comparison only works if everything except the variable stays fixed. Each build starts from one donor car — a specific model, a specific year range, a specific body style — and the donor is never swapped mid-set.
The car is described first: make, model, generation code, year range, body style and the factory proportions that matter (wheelbase, overhangs, shoulder height, glasshouse). Those proportions are the constraint. If a build breaks them, it stops being a build of that car.
Each axis is a short list of permitted changes. Soft Mod may change ride height, wheel diameter, offset and tyre profile — nothing else. Hard Mod may change arches, track, splitter and diffuser. Off-Road may change ride height, tyre sidewall, bumpers, racks and lights. The race axes borrow their rules from the real categories: wheel width, aero devices, ride height and livery placement follow what the discipline allows. When an idea does not fit any axis, it goes to Fantasy rather than bending the rule.
Within a set, the lighting rig, the camera angle and the ground plane stay constant. Only the build changes. That is why the gallery is ordered the way it is: scroll from the first image to the last and the changes accumulate in one direction instead of jumping around.
A render is kept if someone who knows the category would accept it. A rally car needs ground clearance and a light pod, not just stickers. A time attack car needs a splitter that could survive a kerb. A touring car has to stay recognisable as the production model. If the answer is "that is a supercar with a wing", the render is not published.
Builds are filed under car and axis, never under a mood or a date. Every car page lists the axes that exist for it, every axis page shows the complete set, and every style hub collects that axis across the whole library. Images carry the car, the axis and a sequence number, which is what the captions show.
Questions about the method, or a car you think belongs in the library: contact.