Prev | Next
Prometheus Rendering a Greek God
The Prometheus tutorial was created by guest author, Scott Eaton, of Escape Studios in London.
In the Prometheus tutorial, we will use RenderMan for Maya to shade, light and integrate a 3D model into a background plate from the British Museum. During the process, we will explore many of the advanced features that RenderMan for Maya offers including: micropolygon displacements, subsurface scattering, ambient occlusion, brickmaps, and secondary outputs. The tutorial will be broken into three parts:
First, we will develop of a marble shader that approximates the main surface characteristics our reference images. It will need to show small divots from the wear of time on the surface of the marble, discoloration from oxidation, and changes in specularity from polished areas to rough, unfinished areas. And, of course, no marble shader would be complete without subsurface scattering. As the shading network builds up, refer to the Maya materials in the scene files. Each important node will have notes on it to explain its function in the shading network.
Second, we will analyze and match the lighting setup in the British Museum. Without the benefit of an HDRI probe to light from directly, we will have to dissect the lighting from reference photos and an analysis of how sculptures are lit in the British Museum.
Third, we will render out a number of secondary outputs (diffuse, specular, occlusion, etc) with RenderMan for Maya, and use these secondary outputs to build up a composite over a background plate from the museum. This step is where the magic and flexibility of secondary outputs pays off as we layer and correct each output to match the museum plate.
Before and After - In this tutorial we'll add complexity to a simple model using
displacements, occlusion, subsurface scattering, and secondary outputs
This is an advanced tutorial combining a number of different workflows, and because of that, this tutorial requires a good understanding of RenderMan for Maya. You may need to review certain techniques elsewhere. A key strategy of this tutorial is the use of Secondary Outputs, which will allow us to use 2D compositing to finalize our lighting. At the end of the tutorial, the final result should look something like this:
The end result, the 3D model of Prometheus composited over the background image.
1 ANALYZING REFERENCE MATERIAL
Before we get started, open Maya and make sure that RenderMan for Maya is properly set up.
Weathered marble shader reference
1. The surface we are trying to mimic is aged, weathered marble. In our reference image above, we can see that marble is not simply a 3D marble texture mapped onto a Blinn material. The surface is actually a layered combination of cloudy-grey base texture, yellowish surface discoloration from environmental wear, black specs from mineral contamination, faint rain streaks, and whitish marble dust build-up in the deep cavities. Creating a convincing marble will be a challenge.
2. The lighting in the gallery is made up of both direct and indirect light sources. In our reference photo above we can attribute a large amount of non-directional diffuse light to the skylight above the gallery. In addition, each sculpture is individually lit with warm, colored spot lights located near the ceiling.
3. Open the Maya tutorial scene, Stage1.mb, to get started on the tutorial. (Where are the tutorial files?) The initial scene contains a simple light setup and the three parts of our statue: Prometheus, cloth, and chains. Our first step is to add comlexity to this model using displacement maps.
2 SHADING: SUBDIVISION AND DISPLACEMENT
4. We will begin by adding displacement maps to the 3D model which will add the intricacies of a carved marble statue to our lightweight 3D model. Luckily for us, we already have displacement maps for this purpose, originally created using ZBrush. Because it is capable of producing high quality displacement maps, ZBrush is a useful tool for use with RenderMan because RenderMan is capable of rendering these displacement maps extremely efficiently.
Watch UV Seams: When rendering displacement maps it is important to create equal UV seams, otherwise artifacts may appear in the displacement along the UV seams. In the images below, the image on the left has unequal seams, while the image on the right has equal seams.
For the highest quality displacements, we'll want to render the model as a subdivision surface. In RFM, this is as easy as adding a single attribute to the shape node. Select each piece of the statue, one piece at a time, and in the attribute editor select: Attributes-> RenderMan-> Add Subdivision Scheme. Do a test render to confirm that you have a perfectly smooth subdivision surface.
Three Blinn materials with displacements are added to the three parts of the statue
5. Connect three standard Blinn materials to each of the three pieces of the statue, and assign a file texture as a displacement map, as shown above. Get the 32bit displacement maps for each model from the source images directory, sourceimages-> prometheus_tutorial-> [image folders]. In each file node set the Alpha Gain to 20, Alpha Offset to -10. This centers the displacement on 0. A final task: on the file node, turn Filter Type to off. This stops filtering on the displacement map, which can reveal seams.
Displacement mapping: RenderMan for Maya brings RenderMan's extremely fast and accurate displacement mapping to Maya users. Getting used to having all this power takes time and a shift in thinking about geometric detail. Because RenderMan renders displacement mapping as quickly as bump mapping (nearly), there is no reason not to use it extensively in your scenes where geometric detail is required. Think, for example, tire treads on a large truck, a stone wall, etc. All can be represented in your scene with lightweight models but displaced at render time for maximum detail. Experiment by mapping a few procedural textures to a displacement node.
Displacement Attributes are added to the shader and Displacement Bound is set to "20"
6. It is important to tell RenderMan how much you ever intend to displace your surface. This helps the renderer to efficiently process the geometry for rendering. Select each of your displacement nodes and go to Attributes-> Renderman-> Add Displacement Attrs, as shown above. To estimate your displacement bound, roughly use your alpha gain from your displacement map node, in this case "20".
"Trace Displacements" are enabled
7. We must also tell RenderMan that we want to use the displacement mapped version of this object for any ray tracing that we are going to do (occlusion, blurred reflections, etc). To do this we must add another RenderMan attribute to the shape node of our models. Individually select each shape node, of the three parts of the statue, and go to Attributes->RenderMan->Manage Attributes, then select and add Trace Displacements.
Quality Control: A single parameter, Shading Rate, controls the quality of shading and displacement sampling in Renderman for Maya. The control is located in the Render Settings-> Quality tab. The control is simple but powerful. The value is expressed in pixels and equates to the pixel space between shading samples (in screen space) at render time. Therefore, the lower the number, the higher the quality. A value of one, for example, samples textures and procedural networks once for every pixel in the image; similarly a value of five samples every fifth pixel, and is quicker, but lower quality. Consider raising the shading rate to 5-10 to speed your preview renders, but drop it back down to 1.0 or 0.5 for final renders.
3 SHADING: DIFFUSE, SPECULAR, AND SUBSURFACE SCATTERING
The network for the marble shader in the Maya file, Stage2.mb
8. Next, we can build up the patterns from our reference with a combination of procedural textures, high-level control maps, and texture maps. Now we'll take a shortcut and open Stage2.mb and investigate how the marble shader is constructed. Notice that the diffuse, specular, and bump components are built up using layers, many of which are shared between shading components. We will use this network as the starting point for adding RenderMan specific features to our materials.
Shading in Layers: It is good practice to break your shading up into modular layers. If you are building up a diffuse color (either procedurally or by painting maps), build it in components, starting from the largest "low frequency" patterns and work your way to smallest "high frequency" details. By separating these patterns into layers, it makes it easy to change or modify any one component independently. You will also find that some of these patterns will individually effect the specularity, reflectivity, or bump of the material and can be easily reused there. You will see extensive use of Layered Texture nodes in this shading network provided.
Add subsurface scattering attributes to the Blinn materials and adjust as shown
9. Next add RenderMan subsurface scattering attributes to our three Blinn shaders. Select each Blinn material and go to the attribute editor and add RenderMan Attributes-> Subsurface Scattering. This adds a collection of extra attributes to the shader as well as an additional render pass for creating the SSS brickmap (which also shows up in the Render Settings-> Passes tab).
Set the three Blinn's to share the same Brickmap Pass
10. For the SSS pass we want to include all three pieces of the statue into one pass in order to share the brickmap between the three Blinn shaders. In each Blinn, right click over the Scattering Map field and select the first rmanSSMakeBrickmapPass (delete the other SSS passes from the RenderSettings-> Passes tab). Now all three pieces of geometry will be written to and read from one brickmap.
Configure the camera and set for the SSS pass
11. In Render Settings-> Passes, select the rmanSSRenderPass. To pick the baking camera, right-click in the Camera field and select bakingCameraShape, which has already been set up in the scene. To tell the pass which objects to render in the SSS pass, right-click in the Sets field and choose statueSet. (Note: for best SSS baking results, either enable raytraced shadows on lights or have shadow maps being reused, otherwise your SSS pass will not self shadow). Now you're ready to render the effect. Remember that once the SSS effect is rendered, the SSS brickmap can be reused. Since this statue is stationary and the lights are fixed, we'll be able to reuse the SSS brickmap over and over. Next render the effect then switch the Caching Behavior of the SSS brickpass from Compute to Reuse, as shown in the following tip.
Reusing Brickmaps: Bake early, bake often. RenderMan uses an elegant baking system to capture computationally expensive effects into 3D caches called brickmaps. It is important understand that baking is done as a separate render pass, and to get best results you should specify a reference camera in the Render Pass settings. By changing the Caching Behavior of the pass you are able to explicitly control the reuse of the SSS brickmap.
Caching Behavior can be set to Compute, Reuse, or Disable
4 LIGHTING SETUP
Add shadows to the spotlights
12. With the shaders finished, we can now focus on lighting. The scene has two main lighting elements, the spot lights and the sky light. To start, we'll tackle the spot lights. There are four yellowish spotlights that light the sculpture from above. On each of these spot lights, enable depth map shadows.
The shadow pass
13. Next we'll render the scene once to generate all our shadow maps, and then we'll set our spot lights to reuse the shadow maps, so that they are not recomputed every time we render. To reuse shadows, change the Disk Based Dmaps setting on each spot light from Overwrite existing Dmaps(s) to Reuse existing Dmaps(s).
Map Reuse: RenderMan has a long history of reusing pre-computed maps to save on render time for large productions. Each time you create a new pass that needs to compute a map (be it a shadow map, a reference map, or 3d brick map) you get a Phase option to create a disk cache in two ways: once per job and every frame. For stationary, non deforming geometry (as we have in this tutorial) once per job is the option you want to select. For animation sequences where there is motion that will change the map, use every frame and you will get a cached map for each frame of the animation.
14. To give us control over the shadow filtering and samples once we have a shadow mapped baked, we need to add additional custom attributes to our light shape node. Select the light and go to Attributes->Renderman-> Shadow Attributes. This gives us access to the shadow filtering and shadow sampling attributes to blur the edges of our shadow maps as necessary. Now the spot lights are properly set up.
15. Next we'll approximate the skylight in the British Museum. To achieve this, we will combine two effects: an area light for specular highlights and a custom occlusion pass (see next step) for soft shadowing. The area light will be used only to cast accurate specular reflections while the custom occlusion pass will allow us to bake expensive soft shadow calculations. Find the areaSkylight in the scene, unhide it, and enable Emit Specular only.
16. Additionally, in the museum there is light reflected from the floor which is cast from the overhead skylight. To create that effect, unhide the ambient light in the scene ( we will later use the light to attenuate our ambient occlusion pass). Set your intensity low and tune Ambient Shade to 0.2 to give a little direction to our light, to simulate reflection from the ground. The amount of ambient can be tuned later in compositing.
5 OCCLUSION AND SECONDARY OUTPUTS
Set up your environment light as shown
17. To create realistic shadowing, we'll add the global illumination effect, ambient occlusion. To enable global illumination effects, (ambient occlusion in this case) we must create a RenderMan environment light from the RenderSettings-> Features tab. Enable RayTracing and create an Environment Light. The environment light is where we access all our setting for occlusion. Tune the settings as shown, hide your other lights, and render. You should get a classic occlusion result. (If not, make sure your Max Dist is set to "50".) Tune your settings until you are happy.
18. To save recalculating this expensive ray tracing effect every frame, we are going to bake the calculation into a brickmap. First, create a new MakeGlobalDiffuse3d pass in the Bake field of the Environment Light. This will create a new pass in the Render Settings -> Passes tab where you can tune the Caching Behavior and Phase for the pass. We will use compute and once per job, as shown above.
Configure the RenderDiffuse3d pass as shown
19. The other tab created with the new MakeDiffuse3d pass is the RenderDiffuse3d pass, found beside the former tab in the Render Settings. Similar to the SSS setup, this tab is where we setup a camera for final baking. Specify the same bakingCameraShape, and then disable Cull Backfacing and Cull Hidden. Add the RenderMan attribute Dice Rasterorient and disable it. Render to bake, then set Caching Behavior in the MakeGlobalDiffuse3d tab from Compute to Reuse.
Quality vs. Speed: The environment light has a number of parameters for tuning the type, quality and speed of the global illumination effects. MaxVariation controls the amount of interpolation between pixels and is analogous to Shading Rate: try setting this to 1 (or less) for production quality and 10 (or more) for preview rendering.
20. Now we will simulate the illumination of the skylight. We will fake an area light using ambient occlusion, because ambient occlusion makes it easy to bake this data into a brickmap (unlike an actual area light). Hide your first environment light, and unhide the occlusionBlocker mesh in the geometry group. This missing face at the top will act as the skylight letting the occlusion pass see white. Next create new environment light (a second environment light) from the Render Settings-> Features tab by right-clicking on the Environment Light field.
21. We will make this Environment Light act like an area light by setting the Max Distance attribute to a value larger than the maximum dimension of our blocking geometry. Here a value of 10,000 is used. The other settings are the same as the last Environment Light.
22. Use similar bake settings for this pass. Remember to turn off Cull Backfacing, Cull Hidden and Dice Rasterorient and specify the bakingCameraShape. Now render the scene to bake the occlusion brickmap. You should see this brickmap file, along with the other brickmaps, ptc files, and shadow maps is your /renderman/$scene/data folder in your current project (these default directories can be changed in the Render Settings-> Advanced tab).
Add secondary outputs to the Final Pass
23. Now we will create the secondary outputs for compositing the final image. First unhide all the lights, except the fist environment light which should remain hidden. To prepare output passes we go to Render Settings-> Passes tab. Here we add the following secondary outputs to our Final Defaults pass: diffuse (DiffuseDirect), occlusion (OcclusionDirect), ambient (Ambientlight), specular (SpecularDirect), subsurface scattering (Subsurface), & surface color (DiffuseColor). Add these passes by clicking on the black arrow next to Pass and selecting Final Defaults. Next click on the black arrow next to the Output field and select the outputs one at a time. Set all outputs to format Tiff16. Render once in batch mode and look at your secondary outputs, which will appear in the Maya project directory, renderman-> prometheus-> images. To get our original "close occlusion" pass, unhide our first envLight and hide the second "skylight" envLight. Render with only occlusion as a secondary output, and to do that you must delete the other secondary outputs by selecting and deleting them from the pull-down menu. To prevent the first occlusion pass from being overwritten, change its name to occlusion_arealight.tif.
Secondary Outputs: Rendering your scene in separate passes give you maximum flexibility in compositing and greatly reduces the amount of time you need to spend tuning your renders in 3D. RenderMan efficiently generates all the passes you need to rebuild a final render in compositing. Passes are known as secondary outputs in RenderMan, and the renderer is very efficient about outputting these component parts from a single render (any number of outputs are created at one time). Once you start using secondary outputs in compositing and you won't go back. When you break your output down into its component parts: diffuse, reflectivity, occlusion, shadow, etc., you have greater flexibility in color correcting individual layers to more precisely match background plates in hue, saturation, expose, etc.
Layer the secondary outputs with diffuse on bottom
24. The power of rendering with secondary outputs is that it gives you incredible control over tuning the look of your plate in 2D. For this example, we will use Photoshop to demonstrate the basic steps used to composite our secondary outputs, although in a studio compositing might be performed with a dedicated compositing program. For user's of RenderMan Studio, the image tool "it" may be used to perform compositing (see below). Now using Photoshop, open all your secondary outputs and layer them in one image as shown above.
Adjusting lighting in compositing can be much faster than rerendering
25. Now we need to set the blending modes between layers. Diffuse is the lowest layer and above it screen on the ambient channel. Next multiply by the close-occlusion layer. Then either multiply by the area light occlusion to take light out of the image, or screen to add light. Then overlay the surfacecolor pass. Finally screen both the subsurface and specular passes. Use level adjustments to tune the image.
The final composite
26. Finally we color correct our plates to match the background. The main challenge of color correction is finding the correct transition from light to shadow and then matching the color temperature in the highs, mids, and shadows. Here the background plate is very warm red-orange in the mids and highs, while the shadows have more cyan. The final match was done using a combination of color balance for the temperature and levels for transitions.
About the Author: Scott Eaton is currently Creative Technical Director at Escape Studios in London, where he divides his time between production and lecturing. His client list includes Sony, Microsoft Game Studios, The Mill, Double Negative, and many other leading production houses and games studios. Scott received his master's degree from the MIT Media Lab and subsequently studied traditional fine art at the Florence Academy of Art, Italy.
Prev | Next
Pixar Animation Studios