A real-time nose gear assembly built to the standard a simulation or training pipeline actually asks for: modelled to published aircraft dimensions, textured across four PBR sets, cut to LODs with recorded tri counts, and delivered into Unreal Engine 5.
Simulation and visualisation studios do not hire on mood. They hire on whether an asset arrives clean, correctly scaled, correctly named, and cheap enough to run in a real-time scene alongside several hundred others. This piece was built to be judged on those terms. It is the nose landing gear of an Airbus A320, modelled from walkaround reference and published dimensions rather than from a generic gear leg, with the whole pipeline from blockout to engine import documented as it went.
The gear was built against a reference board of walkaround photography rather than a single hero image, so that parts hidden in any one view still had a source. Dimensions were taken from published figures, not estimated from photographs.
The A320 nose gear is free-rolling. Braking is carried entirely by the main gear, so there is no brake unit, no torque plate and no hydraulic brake line on this assembly. Modelling one would have been a visible error to anyone who works on aircraft.
This is the kind of detail that separates an asset built from the actual aircraft from one built from a generic gear leg, and it is the reason the reference stage took as long as it did.
Four texture sets, grouped by how the material behaves rather than by which object it sits on. That grouping is what lets a chrome oleo, a bare actuator rod and a cad-plated fastener share one set without fighting each other.
The four circumferential grooves are authored in Substance as a height-channel fill layer driven by a stripe procedural, masked to the tread band. No tread geometry exists in the mesh at all, which keeps the tire cheap while reading correctly at any distance a trainee would see it from. This only worked because the tire was unwrapped as a clean straight rectangle: a warped or islanded tire UV would have made the stripe procedural unusable and forced the tread into geometry.
Density is consistent within each texture set, enforced by unwrapping each set in multi-object edit mode and running Average Islands Scale. Consistency between the four sets was verified visually with a colour grid checker rather than measured numerically, so no pixels-per-centimetre figure is claimed here. Where a studio needs that number to a spec, it is a measured pass rather than a rebuild.
Four things that were not obvious going in, and how they were solved.
Four levels generated with Unreal's built-in mesh reduction on a halving chain, each checked in the editor for silhouette preservation. Counts read from the engine rather than estimated.
Imported to Unreal Engine 5 with LODs assigned, materials rebuilt from the exported PBR maps, and the assembly checked at the distances a trainee or operator would actually view it from.
The build was documented as it happened rather than reconstructed afterwards: a build plan with the scope and dimensional targets set up front, a running build log, and a handover pack covering naming conventions, texture set assignments, LOD counts and export settings. That documentation is part of the deliverable, because in a simulation pipeline an undocumented asset is an asset someone else has to reverse engineer.