Glass, Metal & Architectural Surfaces

Glass is the surface architects care about most and the one real-time handles most awkwardly. The first decision is the material's blend mode. An opaque material (the default, cheapest) lets no light through. For glass you need light to pass, which means either a translucent blend mode or, for simple cases, dithered/masked opacity. Translucency is what gives true see-through glass with tinting and falloff - but it is markedly more expensive than opaque, and it interacts awkwardly with some real-time lighting and sorting, which is the core trade-off of the lesson.

Good architectural glass is usually less transparent than beginners make it. Real glazing has a subtle tint, visible reflections (the sky and interior reflected in it), a slight roughness, and Fresnel behaviour - it becomes more mirror-like at grazing angles. A pane set to near-invisible clear looks less real than one with believable reflections and a faint green-blue tint. So the craft is balancing transparency against reflection, not maximising see-through.

Refraction - light bending as it passes through glass, so the view behind is displaced - adds realism to thick glass, water and lenses, but it is one of the most expensive and finicky things in real-time. For thin architectural glazing you often deliberately skip refraction (a window pane barely refracts anyway) to save cost and avoid artefacts, and reserve it for thick features where it genuinely reads. This is a place to spend the budget deliberately: know that refraction is costly, and use it only where the eye demands it.

After glass, metal feels intimidating - but it is actually one of the more forgiving hard surfaces, because PBR handles it cleanly once you respect two rules. First, metallic = 1 for raw metal, always (recall lesson 3.1); the moment it is painted or coated it becomes a dielectric at metallic 0. Second, a metal's character lives almost entirely in its roughness: polished chrome is roughness near 0 (a sharp mirror), satin steel around 0.3, brushed aluminium higher and directional, a weathered zinc roof higher still and broken up with variation.

The interesting case is brushed metal, where the tiny grooves run one way, so the highlight streaks along the brush direction rather than forming a round spot. This is anisotropic reflection, and Unreal supports it with an anisotropy input plus a tangent direction. You do not always need true anisotropy - a good directional roughness or normal map fakes brushed steel convincingly for most archviz - but knowing the effect exists tells you why brushed metal looks different from merely rough metal, and when to reach for the real thing (a hero stainless balustrade the camera runs along).

What metal genuinely depends on is something to reflect. A metal surface is mostly a reflection of its surroundings, so a polished balustrade in an empty grey void looks dead - it needs a real environment, an HDRI sky or reflection sources to mirror. This is why metals come alive only once your lighting and environment (Modules 4 and 5) are in place, and why judging a metal material in isolation is a mistake. Give it a world to reflect and the same material transforms.

Reflections are where real-time makes its most visible compromises, and understanding the three main techniques tells you what any reflective surface will actually show. In rough order of cost and quality:

Reflection Captures are cheap, pre-computed probes that store a low-resolution snapshot of the surroundings from a point. They give plausible ambient reflection on most surfaces for almost no runtime cost, but they are static and low-detail - fine for a satin worktop, wrong for a true mirror. Screen Space Reflections (SSR) reflect what is currently on screen, in real time - good and dynamic, but they can only reflect what the camera can already see, so objects off-screen or behind the viewer simply vanish from the reflection, producing tell-tale cut-offs. Lumen reflections (UE5) are the modern answer: Lumen extends its global-illumination system to produce high-quality, dynamic reflections that handle off-screen detail far better, at higher cost - and they are what make real-time archviz reflections finally convincing.

A true, sharp mirror is the demanding case: it needs to reflect the full scene accurately, which is expensive. Lumen high-quality reflections or, for a hero mirror, a dedicated Planar Reflection (which effectively re-renders the scene for that surface - costly, use sparingly) get you there. The practical lesson is to match the technique to the surface: reflection captures for dull reflective surfaces, Lumen for the glossy ones that matter, and a planar reflection only for the one hero mirror that justifies re-rendering the scene. Spraying expensive mirror-quality reflections everywhere is how a scene drops out of VR frame rate.

It is worth returning from the glamorous surfaces to the ones that make up most of a building, because they are where correct PBR pays off most reliably. Concrete is a masterclass in roughness and normal detail: it is a dielectric (metallic 0), fairly rough (0.7-0.9), and lives or dies on subtle variation - the faint sheen where it has been polished, the roughness change where it is stained, the normal-mapped board-form texture or aggregate. Flat, uniform concrete is the fastest tell of a lazy scene; broken-up concrete is deeply convincing and cheap to render.

Wood rewards the same care: a dielectric with a roughness that varies between a sealed gloss floor (low) and an oiled matte tabletop (higher), a strong normal map for grain, and often a subtle anisotropy or directional highlight along the grain that lifts it. The base-color texture carries the timber's figure; roughness carries whether it is polished or raw; the normal carries the grain you feel. Get those three honest and wood is one of the most satisfying surfaces in archviz.

The throughline is that these everyday surfaces are cheap and forgiving - opaque, no refraction, standard reflections - and they benefit enormously from the fundamentals of this whole module: correct PBR values (3.1), a clean instanced material (3.2), and true real-world texture scale (3.3). Spend your exotic-surface budget - translucency, anisotropy, planar mirrors - only where glass and hero metal genuinely demand it, and keep the bulk of the building on solid, efficient opaque materials. That balance is what keeps a richly-detailed scene comfortably interactive, in a flat walkthrough and especially in VR.

Pulling the module together: reflective and transparent surfaces are where realism and performance most directly collide, so they demand deliberate budgeting - never more true than in VR, where you render the whole frame twice and cannot afford a surface that halves your frame rate. A few honest rules keep you safe.

Translucency is expensive; use it where glass truly needs it and prefer opaque everywhere else. A whole facade of true translucent glass will cost you; often a well-reflective, slightly-tinted glass with reflections doing the work, and refraction switched off, reads beautifully for far less. Refraction is a luxury - reserve it for thick glass, water and lenses, not every pane. Match the reflection technique to the surface - captures for dull, Lumen for glossy heroes, planar for the single mirror that earns it. Give metals and glass a real environment to reflect, or they look dead no matter how well authored.

And remember the honest limit from Module 0: real-time approximates these effects. A single uncompromising hero still of a glass tower at golden hour may still be worth an offline render where accuracy beats interactivity. But for the walkthrough, the client review and the VR walk - the things this course exists for - well-judged real-time glass and metal are now genuinely convincing, provided you spend the budget where the eye looks and keep the rest of the building on efficient opaque materials. That judgement - what to splurge on, what to keep cheap - is the real skill this module has been building toward.