I notice that most photorealistic renderers have very similar material test scenes. Here are a few examples:




My question is: why is this specific model / style chosen to test materials? Does it have some advantage over teapots, spheres, and suzannes? Are there other common scenes for material testing?


While teapot, spheres and trusty Suzanne are not per se bad test scenes for materials, here are some things to consider. You can apply them to said examples to forge your own opinion.

First and foremost, the viewer needs to be able to examine the behavior of the BRDF. Since it is dependent on both the direction of incident light and the direction from surface to viewer, you need to have as many combinations as possible in your test scene. A spheroid nicely covers all possible viewer angles.

Notice that all three scenes feature some sharper edges or creases too. Differences in reflective behavior across a smooth surface (like a sphere) may be hard to conceive, so these edges put emphasis on this.

Since the observed brightness and color is a combination of the material (i.e. the BRDF) and the lighting environment, it may be hard to tell one from another. The Vray test scene takes care of this. The 100% white 'core' and the 25% black floor help you get an idea of what the lighting in the scene is like.

Blender and Vray both feature a grid that helps you estimate the total size of the scene. This helps with evaluating grain or surface texture.

Here is a side-by-side comparison of a cube, notched sphere and Suzanne for a glossy test material with a slight normal map (rendered in Blender with Cycles, lit with the Dutch Skies 360 - Free 002 HDR):

Material test scene comparing cube, spheroid and Suzanne

The cube does a pretty good job on depicting the normal map. The sphere, as expected, is reasonably good for normals and gloss. Suzanne, arguably, makes both a bit harder to perceive than the spheroid.

  • $\begingroup$ Yeah, complex surfaces leave something to be desired if the textures need to be appraised, as you can not distinguish easily what is texture and what is not. In this case Suzanne is also badly oriented as its predominantly showing the flat angle towards the camera. $\endgroup$ – joojaa Sep 17 '15 at 9:34
  • $\begingroup$ @joojaa You're right, the main reason why Suzanne loses here is probably because of the angle. I'll mention that in the answer, or better, render another pic with better orientation. $\endgroup$ – David Kuri Sep 17 '15 at 10:31
  • $\begingroup$ Looks better, still complex shapes are harder to appraise now i dont know if certain features are because of the normal map or are they there because of Suzannes shape. PS: maybe the literal answer to the question should be "because they are trying to solve same problem" $\endgroup$ – joojaa Sep 17 '15 at 11:35

You can choose whatever you want offcourse. A good material sample ideally has a few properties.

  • It needs to show how the light falls of as the angle between light and surface normal changes.

    A sphere is pretty good for this.

  • It needs to show how any possible environment, maps, reflects, refraction etc.

    A sphere is pretty good for this.

  • It may need to show how nonlocal effects happen such as global lightning, incandescence, subsurface scattering work

    Some cavities and variation would be nice. A sphere is abysmal for this.

  • The sample should be able to show how displacement looks like.

    A simple shape demonstrates this better. Complex shapes can cause displacements to self intersect. And its hard to know what is modeling and what is not.

Also it would be nice to see, different surface thicknesses, sharp corners, one way curved surfaces and doublecurvature surfaces. A neutral but nonuniform environment would be nice too. Object should have a reasonable uv map. Apparent scale might be needed in some projects.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.