What Is PBR? Physically Based Rendering in 3D Materials

ArticlesMaksim Bronnikov
img00.png

What is PBR and why does it matter for 3D artists? In short: it is a way to store surface properties separately from lighting — so the material behaves predictably in any scene. Let’s break it down with a simple example.

Take an ordinary apple and look at it by a window, at sunset, or under a desk lamp. In all three cases, it will look different: brightness, hue, highlight position, and shadow depth will change. But the apple itself has not changed — only the way the surface appears under that lighting has changed.

We intuitively understand how the same object should look under different light sources: how its color and brightness will change, how much light will be reflected, and how much will be scattered. We don't need to know physics — if something does not match our expectations, we notice it immediately, even when the shape and color of the object look convincing.
PBR tries to describe the properties of an object so that, in a virtual world, the material matches that expectation.

PBR (Physically Based Rendering) is an approach to rendering and material description in which the interaction of light with a surface is modeled using simplified approximations of real-world physics, rather than adjusted by an artist “by eye” for a specific scene. In practice, the main idea is simple: the material should work under different lighting.

Google, in the technical description of Filament, calls separation the foundation of the method:
“The separation of materials and lighting at the core of the PBR method makes it easier to create realistic assets…” — Google Filament, Physically based rendering

The same apple under cool window light, a warm desk lamp and blue screen light.
One material, three responses to light: the frames differ, yet all three look believable.

What Exactly PBR Separates: Material and Lighting

To understand what exactly needs to be separated, imagine a photograph of an apple. Everything is mixed together in it: light, shadows, surface color, reflections. From the photo alone, it is no longer possible to tell what belongs to the material itself and what was caused by the environment and capture conditions.
PBR textures break this mixture into separate properties:

  • what color the surface itself is;
  • whether it is smooth or rough;
  • whether it is a metal or a dielectric — wood, stone, plastic, ceramic;
  • whether it has additional properties such as transparency, emission, or subsurface scattering, and many others.
Apple photo with shadow and highlight beside its Base Color, Roughness, Metallic and Normal PBR maps.
A photo is one finished result; a PBR material is a set of separate properties.

A photograph is one finished result. PBR is a set of separate properties.

How PBR Works: The Four Participants in an Image

But a set of properties is not an image by itself: for an image to appear, the material properties must interact with the scene. That is why each pixel should be broken down into four groups of causes.

Participant What it determines Everyday example
Geometry The shape and orientation of the surface A mug handle casts a shadow; a dent changes the angle
Material How the surface reflects, absorbs, and transmits light Glass, wood, and steel respond differently to the same lamp
Lighting Where the light comes from, what color it is, and how intense it is Window, candle, sun, LED panel
Camera How light turns into an image Exposure, white balance, tone mapping

image = geometry × material × lighting × camera
The multiplication sign is symbolic; the real relationships are much more complex. But the main point is what they show: the brightness of a pixel cannot automatically be called the color of the material. A dark pixel may mean black paint, shadow, low exposure, or all of these at once.

Apple mesh, texture maps, a studio light and a camera combining into one final apple image.
Every pixel has four causes, so its brightness cannot automatically be read as the material's color.

“Works Under Any Lighting” Does Not Mean “Looks the Same”

One point is worth clarifying just in case. The phrase “a PBR material works under any lighting” does not mean that an object will look the same under all conditions.

A white T-shirt under a red lamp will become red. A black boot in an almost dark room may disappear. A polished spoon in an empty black room will look almost black because it has nothing to reflect. A more accurate way to say it is:

When the light changes, the image changes, but the surface properties do not.

The difference is easier to see in comparison.

Pear with baked lighting and with a PBR material, lamp moved from left to right.
Light baked into the texture contradicts the new light; the PBR material responds to it consistently.

Pear A: lighting is baked into the texture. Under an unchanged environment, everything looks fine. But move the lamp to the right — the renderer will light the other side, while the old shadow remains in the texture. The right side will be both lit and dark at the same time, and the left shadow side will be bright. The brain notices the contradiction even if you do not know the term
PBR.

Pear B: lighting is separated from the material. Base Color stores the surface color without the old shadow, Normal describes fine relief, and Roughness describes the character of the reflection. When the lamp moves, a new lighting pattern appears where it should. One material produces many different, but causally consistent, frames.

Pear A is not a fictional example, but common pre-PBR practice: lighting was baked directly into the diffuse texture because calculating it in real time was too expensive.

  • the material looked believable only in the scene and under the lighting it was prepared for — in another environment it looked “flat” or had inappropriate highlights;
  • the material could not be predictably reused between engines and renderers;

A PBR material under different lighting should not look identical; it should react consistently to the new light.

Which Maps Make Up a Typical PBR Material

The de facto standard today is the Metallic/Roughness workflow: it is the default in Unreal, Unity, glTF, Substance 3D Painter/Designer, and Blender Principled BSDF. The minimal set is Base Color, Metallic, Roughness, Normal, and AO.

Map What it stores
Base Color / Albedo The intrinsic surface color without baked shadows or highlights: for dielectrics, the color of the diffuse component; for metals, the color of the reflection itself
Metallic 0 = dielectric, 1 = metal; almost always binary in meaning
Roughness The width of the specular response: 0 means mirror-smooth with a sharp highlight, 1 means a maximally broad, blurred reflection
Normal The normal direction per pixel — an imitation of fine relief without changing the mesh
Ambient Occlusion An approximation of self-shadowing in pores and crevices from diffuse ambient light

The parameter set is intentionally small. When creating the principled model, Disney chose “intuitive rather than physical parameters” and at the same time required “as few parameters as possible” (Brent Burley, Physically Based Shading at Disney, p. 11) — the point is not to expose all of physics to the artist, but to provide a small and stable set of controls.

Each map also has a technical side: color maps store perceived color and require sRGB→linear conversion when read, while all the others store technical values that are read without gamma conversion. Confusing these two remains one of the most common import mistakes.

How These PBR Maps Work Together

PBR materials work as a single system: no single map creates a complete image on its own — only their combined contribution to the calculation matters.

Normal defines where the surface is pointing at that point. Metallic decides which path the light takes next: in a dielectric, part of the light reflects from the surface, and part penetrates inside, scatters, and returns colored — this is the diffuse component whose color is stored in Base Color. In a metal, the penetrated light is quickly absorbed, the familiar diffuse component is practically absent, and Base Color defines the color of the reflection itself.

Roughness determines how sharp or blurred the reflection will be. On a smooth surface, the microfacets are oriented almost the same way, so reflected light creates a sharp highlight. On a rough surface, they point in different directions, and the highlight becomes blurred.

Red dielectric and gold metal spheres, smooth and rough red spheres, each with a light-ray cross-section.
Metallic decides where the light goes; Roughness decides how blurry the reflection is.

On top of this, two rules operate that are not stored in any map. The first is Fresnel: the amount of reflected light depends on the angle between the view direction and the normal, so almost any surface becomes more mirror-like at grazing angles. You can easily see this on a lake: look down and you see the bottom; raise your gaze along the surface toward the horizon and you see a mirror reflection of the sky. The properties of the water are the same; only the angle at which we look at the surface has changed.

Clear lake water seen from above, showing pebbles, and at a grazing angle, mirroring the sky.
Fresnel: the share of reflected light grows toward grazing angles, though the water stays the same.

What PBR Does Not Mean

From the definition of PBR — physically based rendering — we understand that the word “physical” in the definition should not be taken literally. It is physically based, but not fully physically correct. Production models do not solve Maxwell’s equations and do not guarantee a pixel-perfect match with a photograph; Disney described this boundary honestly:
“Our philosophy has been to develop a ‘principled’ model rather than a strictly physical one.” — Brent Burley, Physically Based Shading at Disney, SIGGRAPH 2012, p. 11

PBR is not a synonym for photorealism: a physically based material can be used perfectly well in a stylized game, while a photographically convincing frame may be assembled from hacks that fall apart when the light changes. PBR is not a synonym for scanning: a material can be created manually by an artist, assembled procedurally, obtained from photogrammetry, measured, or predicted by a neural network.

The latter means that the mere presence of maps proves nothing. A file named roughness.png may be a fitted parameter from multi-light observations, an artist’s drawing, a grayscale copy of a photograph, or the output of a neural network — and all four options may look equally plausible in a preview.

How PBR Relates to 3D Scanning and Photogrammetry

Ordinary photogrammetry reconstructs the shape of an object well, but the color texture is usually built from photographs. Along with color, shadows, highlights, and other traces of the lighting at the shooting location can get into it. Such a model looks correct either under similar conditions or after significant manual cleanup.

ScanWOW solves this task through controlled capture. The object is captured from different angles and under directional lighting, the camera and lighting are calibrated, and polarization helps separate the diffuse component from the specular component. Then photogrammetry and photometric stereo reconstruct the geometry and surface properties, after which the material is tested under new lighting.

That is why the result of ScanWOW is not a photograph stretched onto a mesh, but a set of separated PBR data that can be transferred between scenes.

Top-view diagrams: camera circling an apple under one light versus several polarized lights.
Ordinary photogrammetry changes only the camera position; controlled capture also changes the light direction, and filters separate the highlight from the color.

The ScanWOW library contains scanned assets created according to this principle: separated PBR data that can be transferred between scenes and engines. Open the catalog of scanned PBR assets.

Key Takeaways

PBR is not just a set of maps and not a way to automatically make a render realistic. Its main task is to store surface properties separately and calculate separately how light affects them.
That is why a correct material under different lighting should look different. Highlights, shadows, brightness, and light color change, but the material properties themselves remain the same.

For 3D scanning, this leads to a very practical conclusion: it is not enough to get good geometry and stretch photographs over it. The lighting captured in the source images should, as much as possible, be separated from the properties of the surface itself. Otherwise, together with the model, we preserve not only the object, but also the conditions in which it was captured.

FAQ

What Is PBR in Simple Terms?

PBR (Physically Based Rendering) is a way to store surface properties separately from lighting and calculate their interaction using physically motivated rules.

What Are PBR Textures?

PBR textures are image maps that store separate material properties, such as base color, roughness, metallic, normal detail, and ambient occlusion. Together, they tell the renderer how a surface should respond to lighting.

Why Should a PBR Material Work Under Any Lighting?

Because the material stores only surface properties, not the result of specific lighting. Light and shadows are calculated separately, so the material can be moved into another scene — and it will respond predictably to the new conditions.

How Many Maps Does a PBR Material Need?

A simple opaque material usually needs Base Color, Roughness, Metallic, and Normal. Glass, fabric, skin, or lacquer will require additional parameters.

Can You Make PBR From a Single Photograph?

No, but you can get a plausible estimate. It is impossible to unambiguously recover all independent properties. A photograph mixes light, material, shape, and camera. Additional angles, known light source characteristics, and polarization reduce the ambiguity.

What Does Physically Based Rendering Mean?

Physically Based Rendering means that the renderer uses simplified physical models to approximate how light reflects, scatters, and is absorbed by surfaces. It is physically based, not perfectly physically accurate.

Sources

  1. Google Filament. Physically Based Rendering in Filament.
  2. Burley, B. Physically Based Shading at Disney. SIGGRAPH Course Notes, 2012.
  3. Pharr, M.; Jakob, W.; Humphreys, G. Physically Based Rendering: From Theory to Implementation, 4th ed.. 2023.
  4. Practical Physically Based Shading in Film and Game Production. SIGGRAPH 2012 course.
  5. Cook, R. L.; Torrance, K. E. A Reflectance Model for Computer Graphics. ACM Transactions on Graphics, 1982.

Ещё статьи