Pre-face
The use of computer graphics to speed up work and/or improve visual fidelity in animation has always been a goal. Rendering has acted as the medium and the in-between for many years, shifting from very flat drawings to full emulations of reality. In general, rendering refers to any process of taking collections of data, whether polygonal, NURBS surfaces or another set, and converting them into the visual idea of real or three-dimensional objects.
Introduction to Rendering
Most often, we 3D artists see systems called scanline or rasterizing renderers. This is a baseline engine designed to draw objects as simply as possible. Taking in object textures, borders, and order (the raw distance of an object and not its face from the viewer or camera), these, today, will often act as the base-line between the artist and the graphics software. A good example to think of this is Maya Hardware 2.0, also called Maya Viewport Renderer. Shading generally is rudimentary, based more on surface quality than any existing light source.
The next step up in complexity in terms of rendering is ray casting. This takes rasterizing a step further by calculating per-face or per-normal lighting data based solely on angle of incidence, the angle at which a ray (light in this case) hits an object. Most often, this value is combined with rasterized data by multiplying the energy per pixel with the full blast value from the scanline or raster render. This gives objects more believable volume as it won’t have the turntable effect, where it appears the lights are static and the object rotates, or the flattening effect, where objects appear flatter or paper-thin due to uncomplex or no shading.
At this point in time, we have reached the final part of rendering. Ray tracing is just ray casting just several times. Rays of lights do not end once they hit an object in real-life, and this is the direction ray tracing takes. Instead, rays continue until the drop below a certain energy level, go a certain distance from the viewer or camera, or after a certain number of bounces. Ray tracing has progressed from suggesting the idea of a surface to portraying near realistic volumes.
An important note is most modern renderers are what are called physically based renderers, where data relates directly to real-life data. Even more artistic aspects of these renderers are dictated by realistic data.
Comparison between Pixar RenderMan and Solid Angle Arnold
Because both RenderMan and Arnold are physical based renderers, many things will appear the same to an untrained eye. Other than names, attributes between their basic surface shaders are mostly the same. RenderMan has a few more lobes, but many of these properties can be emulated within Arnold using utility nodes or utility shaders. However, one notable difference between the two is the treatment of rays. Arnold attempts to emulate reality as closely as possible by treating each ray as the carrier of three kinds of data: hue, intensity, and saturation. This allows Arnold to create chromatic aberration banding like for a curved glass or a carved diamond. Compared to RenderMan’s three color system, it takes longer to render but appears much more physically accurate. As well, since Arnold is designed with current render packages in mind, it can better utilize the nodes within its supported packages.
Comparing these renderers, I personally prefer RenderMan due to its higher level of built-in artistic controls and improved overall performance. I do not fully understand all the smaller details that effect their performance, but due to their similarities, the choice of renderer is purely to the studio’s preference.
Learning the lobes
Diffuse/Base – In general, diffuse makes up the majority of the color of a dielectric (non-metallic) surface. Most often in computer graphics, materials use Lambertian diffusion, where roughness due to nanoscopic structures is consistent. This creates a soft border around the object of grey. However, another important model is the Oren-Nayar or OREN model. Instead of just a consistent nanostructure, Oren and Nayar added a microstructure of V-shaped facets to increase the roughness without greatly increasing computation time. This gives large bodies like the Moon or Mars a more static border. You can read more about these here (at ScienceDirect).
Specularity – Compared to diffusion, specularity makes up the majority of the color of a conductive (metallic) surface. The shift to physically-based renderers has pushed out empirical models, where it is aligned with what is seen and not what happens, but I feel that the understanding of both is important.
The Phong model describes the rate of decay in the intensity of a specular highlight. Phong calculated that f(θ,n)=cos^n (θ), which states the result of the cosine of angle theta is taken to the power of n, such that the higher the value of n, the more true reflection takes place. This is similar to both index of refraction and roughness wrapped into one, as it describes visually what occurs. You can learn more about this here (at Siggraph).
In physically-based renders, commonly index of refraction and roughness are two separate attributes. Index of refraction (IOR) is the ratio of the speed of light through a vacuum to the speed of light through the medium, which affects the angle of reflection and potentially aberrates the resulting ray when changing mediums with varying values. Higher IOR will appear more chrome-like. Roughness then describes how the reflection is broken up, as its name suggests. You can learn more here (at Britannica).
An important note is the Fresnel model, which is a related bridge between IOR and Phong Cosine Power. Fresnel Shape is directly related to, but not, IOR. One can calculate one with the other using the following math equations: R_i (f_0 )≅1+f_0^(1/π) and f_0 (R_i )≅((〖1-R_i)〗^2)/〖(1+R_i)〗^2
Coating – Similar to the specularity lobe, coating or clear coat describes a supsurface volume covering the surface in question. Used in conjunction with specularity, a number of materials can be created with extreme accurancy.
Glass Refraction – Also similar to specularity, this describes light travelling through a volume similar to glass or water. This lobe or setting generally cannot be used with other lobes in a believable manner.
Surface Quality Modifiers – Normal, bump, and displacement change the appearance of the surface in one form or another. Normal maps due this by changing the angle of incidence on a surface to increase fidelity without significantly increasing render-times. Bump and displacement maps generally act as a height placement on the surface. Bump becomes a normal-like, not modifying the actual geometry, just the point at which light hits. While with displacement, geometry is baked and expanded to match the map in question.
Using the lobes to their fullest
Eyes can be hard to reproduce, being both surface and volume like. Here I used Substance Painter and RenderMan to reproduce an eye with only one hour of work.
First, we create the geometry of our eye, as well as the normal map volumes we will need to bake.
Next, we create the base material of the cornea and the pupil.
Then, we create the water coating around the eye.
Finally, we connect our files into the shader as follows.
Now, we have our result.
