By IAN FAILES
By IAN FAILES
Launched in May 1988, Pixar’s RenderMan is celebrating its 30th anniversary this year. The renderer, which Pixar uses internally and licenses commercially and non-commercially, has been one of the mainstays of computer animation and visual effects production over the past three decades, with its origins in computer graphics going back even longer.
Today, RenderMan is a household name and among a select group of production renderers used to produce photorealistic computer-generated imagery. To look back on 30 years of the software, VFX Voice caught up with some key Pixar personnel, who also dove deep into the archives for a mix of old school and new Renderman-related imagery.
Before RenderMan, there was the Reyes scanline rendering algorithm. Reyes (which stands for ‘Renders Everything You Ever Saw’) was developed in the mid-1980s by Loren Carpenter, Robert L. Cook and Ed Catmull at Lucasfilm’s Computer Graphics Research Group, an entity that would later become Pixar. Essentially, RenderMan is an implementation of the Reyes algorithm. The first film to use RenderMan was Tin Toy (1988). A year later, Pixar released the first commercial version of the software.
The name ‘RenderMan’ was coined out of a conversation that founding Pixar employee Pat Hanrahan had about futuristic rendering software that was so tiny it could fit inside a pocket, a la Sony’s Walkman. RenderMan became somewhat of a household name when Pixar’s Toy Story, the first feature-length computer-animated film, was released in 1995 (at the time it took an average of seven hours to render one frame for the film in the renderer).
Like other renderers, RenderMan is typically in a state of continuous development. It has had 27 major and 56 minor public releases in the past 30 years, with notable extensions enabling ray tracing for accurate specular reflections and shadows, ambient occlusion, distribution ray tracing, and point-based global illumination. More recently, RenderMan was re-written as a path tracer via a new ‘RIS’ architecture, as principal software developer Per Christensen describes.
“Instead of the programmable shaders written in RSL, the old school RenderMan Shading Language that handled both material properties and controlled the tracing of rays, with RIS, we split the shaders into programmable integrators to handle light transport (ray tracing and global illumination), Bxdfs calculating surface reflection, and patterns to define surface parameters such as diffuse albedo, specular reflection and transmission coefficients, and transparency. This split was done to better support physically-based materials and realistic light transport. A large chunk of development work went into defining an efficient public interface between these, and writing Integrators for path tracing, and Bxdfs for different material types.”
RIS was introduced in RenderMan 19; the renderer is now at RenderMan 22. Pixar’s Coco (2017) proved to be a major exemplifier of the capabilities of the new path-tracing approach, especially for the vast towers seen in the film’s Land of the Dead which were covered in thousands of practical lights.
“Some of these are compound lights, that is, each single one, represents a large cluster of additional lights,” observes Pixar CTO Steve May. “If you expand out those compound lights, there are scenes that have over eight million individual light sources. Originally, the most ambitious sequences set in the Land of the Dead were completely un-renderable. Through a lot of effort on both the RenderMan side and from Coco production we were able to optimize and provide new features in RenderMan that made the creation of those scenes possible.”
“Many people will have seen R&D presentations from principal engineers and project leads such as Per Christensen, Christophe Hery and Max Liani, or entertaining and in-depth product introductions from marketing manager Dylan Sisson. The core engineers also participate at all levels in events like SIGGRAPH, even staffing demo pods themselves for one-on-one conversations, as well as contributing directly to customer support.”
—David Laur, Business Director, Pixar
The team members behind RenderMan are located at Pixar’s main studio in Emeryville, California, and in Seattle and London. With Pixar (a Disney-owned company) the RenderMan team regularly collaborates with Industrial Light & Magic (ILM), Disney Animation and Disney Research on rendering technologies and techniques. Several RenderMan developers have more than 20 years of experience working on the software and, according to business director David Laur, have a unique tradition of representing their work personally.
“Many people,” says Laur, “will have seen R&D presentations from principal engineers and project leads such as Per Christensen, Christophe Hery and Max Liani, or entertaining and in-depth product introductions from marketing manager Dylan Sisson. The core engineers also participate at all levels in events like SIGGRAPH, even staffing demo pods themselves for one-on-one conversations, as well as contributing directly to customer support.”
Indeed, RenderMan does have a strong ‘aurora’ in the CG community, as evidenced by their presence at SIGGRAPH where they hold their own ‘Art & Science Fair.’
“Since its inception, RenderMan has been focused on creating the most visually rich and complex images possible with the currently available computing hardware. That has always been essential to Pixar, and to RenderMan’s customers, who often produce the most demanding and high-end film content. With the immense growth in computational power and improvement in rendering algorithms, we are now within reach of providing all of that visual complexity at interactive speeds for our artists.”
—Steve May, CTO, Pixar
“The Art & Science Fair is our biggest event of the year,” notes Sisson, “where we usually reveal our major release and show some of the crazy things people are doing with RenderMan at Pixar and also at other studios like MPC and ILM. RenderMan has a broad toolset, and it’s always interesting to see what people are creating with it. I suppose you have to be a little bit of a rendering geek to appreciate it, but there are worse things to do with yourself.”
Sisson, who hosts the Art & Science Fair, is also behind an iconic RenderMan artifact: the RenderMan Walking Teapot. Since 2003, the teapot has been given out each year at SIGGRAPH and is now a major collectible among CG enthusiasts. Its existence harks back to the Utah teapot, one of the first 3D models, created in 1974 by Martin Newell. “Naturally, it appeared everywhere, like SIGGRAPH papers and as Easter eggs in many 3D packages,” says Sisson. “So it seemed fitting to turn it into a windup toy, sort of an insider joke for CGI.”
“The Art & Science Fair [at SIGGRAPH] is our biggest event of the year “where we usually reveal our major release and show some of the crazy things people are doing with RenderMan at Pixar and also at other studios like MPC and (Industrial Light & Magic (ILM). RenderMan has a broad toolset, and it’s always interesting to see what people are creating with it. I suppose you have to be a little bit of a rendering geek to appreciate it, but there are worse things to do with yourself.”
—Dylan Sisson, Marketing Manager, Pixar
Each new film made by Pixar and each new project crafted by VFX studios and other animation studios that rely on RenderMan tends to push the software forward. And whereas traditionally the renderer has been used to produce ‘picture-perfect’ final film frames that could take hours to render, RenderMan 22 has now also implemented an interactive path-tracing approach, i.e. allowing for the renderer to draw the interactive viewports in applications such as Maya, Katana and Pixar’s own animation toolset, Presto.
“Our goal was to make RenderMan the renderer of choice for interactive settings such as geometry modeling, animation, scene layout, object texturing and scene lighting,” explains Christensen. “For example, we want a modeler or animator working in Maya or Katana to visualize the object they are creating using RenderMan from the very beginning rather than using, for example, Maya’s built-in renderer at first and only switch to RenderMan as a last check at the end.”
Other key developments have also been made in recent incarnations of Renderman, including interactive editing of objects, optimizing curves for hair and fur, and, says Christensen, “continued work on improving progressive and adaptive sampling, as well as volume illumination and scattering.”
“Since its inception,” adds May, “RenderMan has been focused on creating the most visually rich and complex images possible with the currently available computing hardware. That has always been essential to Pixar, and to RenderMan’s customers, who often produce the most demanding and high-end film content. With the immense growth in computational power and improvement in rendering algorithms, we are now within reach of providing all of that visual complexity at interactive speeds for our artists.”