• On RenderMan 11 - Interview with Dana Batali from Pixar
    Ali Tezel & Leonard Teo, 27 August 2002
     Image: Monsters, Inc. was the first film to make use of PRMan 11's Deep Shadow technology for Sulley's fur.

    Pixar’s Photorealistic RenderMan (PRMan) rendering system has come a long way since its inception in Lucasfilm’s 1982 motion picture “Star Trek II: the Wrath of Khan”. The ubiquitous renderer is now staple for almost every major visual effects studio in the world for its speed, stability and efficiency. In the last ten years, eight out of the ten films that won the Oscar for Best Visual Effects used PRMan and the creators Rob Cook, Loren Carpenter and Ed Catmull were awarded the 2000 Oscar for their “significant advancements to the field of motion picture rendering as exemplified in Pixar’s RenderMan”.

    Age is catching up with the software though, and while RenderMan maintains its performance and stability for time-critical productions, its lack of advanced features such as global illumination and raytracing has been of concern lately. With RenderMan Release 11, Pixar intends to change all that with the introduction selective raytracing such as reflections and refractions, indirect illumination such as GI, caustics, hemispherical lights, and “Deep Shadow” technology for rendering realistic shadows on micro-objects such as hair. 3D Festival’s Creative Editor Ali Tezel interviews Dana Batali of Pixar regarding the new release of RenderMan.

    3D Festival: What are the key highlights of RenderMan Release 11?

    Dana Batali: We have a new feature called “Deep Shadows”. Shadow maps are wonderful as you’re able to render shadows quickly but this technique has limitations related to aliasing and artifacting. Simply because shadow maps are much faster than other shadow algorithms, we use them all the time. Deep Shadows manages to solve most of the problems associated with shadow mapping such that we can obtain the features not found with regular shadow mapping without having to pay (in terms of render time) with raytracing or other fancier techniques of shadow generation.

    Deep Shadows are particularly good for any kind of blur effects -- we can actually render motion blur with our shadows. We can now also render translucent effects and colored shadows. The most important improvement though, is that Deep Shadows works extremely well with tiny pieces of geometry such as hair.

    For “Monsters, Inc.”, we used Deep Shadows for the hairy characters such as Sulley. We had to get the shadows perfect and this was all done with Deep Shadows -- no raytracing. We developed the technology during production of “Monsters, Inc.” and are finally releasing it commercially as part of Release 11.


    3D Festival: Given that global illumination isn’t used often in time-critical productions, what was the reason for implementing it in RenderMan?

    Dana Batali: For animation, brute force global illumination is still far too slow. Even our implementation it’s still slow for production use. With processor speeds increasing constantly though, it’s good to have the technology in place. Global illumination is going to play a larger role in the near future as the advances in technology makes it feasible for widespread use. The implementation of GI is simply to get our customers going until that time.

    Our global illumination is based on two core technologies: raytracing and photon mapping. In the process of adding the core raytacing functionality, we are able to render refractions. In addition, we are able to use photon maps for the caustics. To get these kind of effects with RenderMan tricks is difficult, and that’s what makes the new functionality great – you can obtain these effects accurately and quickly. It’s the combination of traditional raytracing technology in the RenderMan Shading Language and the REYES (PRMan) architecture that makes it an amazing combination.

    Programmers and shader writers now have access to new Shading Language functions. While standard raytracers simply gather color information from surfaces, our hybrid raytracing/REYES technology allows us to gather anything. We call this “Ray-based Messaging” as it allows us to gather information about any attribute of a surface -- for example, how hairy a surface is. Because you can gather arbitrary information, your shaders are able to do rather fancy things.

    3D Festival: Are there any speed increases between the last version and Release 11?

    Dana Batali: Absolutely. We have significant core performance gains. Independent of all the new features, one customer reported up to 300% speed increase for one of their shots. We are, however, seeing typically between twenty to fifty percent speed improvements for very complicated shots. A single sphere isn’t going to render any faster, but a million or ten million hairs will render faster.

    Next page (2 of 2) >>

  • 3D Festival: Are there also any upgrades to MTOR (Maya to RenderMan connection)?

    Dana Batali: We added new functionality to control the new features in Release 11. This includes new RenderMan globals to turn on raytracing, and control all the variables for photon mapping.

    As far as shaders are concerned, there are new templates that allow you to build shaders for selective raytracing. For example, one shader has a new section for refractions and reflections. With this, you can turn the strength down to zero such that there is no reflections or raytracing. By adding a texture map to control this, non-black areas will have raytracing. The RenderMan Shading Language development system called “SLIM” allows you to connect all these new raytracing parameters to your shaders. This is by far the easiest way to create RenderMan shaders. [3DF]

    Related Links
    Pixar RenderMan

    Images courtesy of Pixar&Disney.

    << Previous page (1 of 2)Back to 3D Festival

blog comments powered by Disqus