Advanced_Renderman_Book[torrents.ru]

Preface

This book explores the theory and the practice of using RenderMan to create computer graphics imagery (CGI) for use in motion pictures and animation. We have written many papers and taught many courses on the use of RenderMan, and this is the most advanced material we have presented yet. Readers are expected to understand the basics of using RenderMan. In this book, we will explore more advanced details on how to define scenes, discuss how to write truly excellent shaders, and will spend quite a bit of time revealing the tricks that are used in real motion picture production to create computer graphics animation and CGI special visual effects. Most importantly, we will explore topics that are beyond the scope of the only previously available reference book The RenderMan Companion.

Let us admit up front the primary bias of the authors of this book. We are the designers, implementers, and users of two specific RenderMan-compliant renderers: Pixar's PhotoRealistic RenderMan (referred to in this text as PRMan) and Blue Moon Rendering Tools' rendrib (referred to in this text as BMRT). These renderers are not the only renderers that have ever attempted to support the ambitious RenderMan Interface Specification, but they are the most well known and the most widely used. In fact, PRMan is so generally considered the "reference platform" for RenderMan that the word "RenderMan" is often used synonymously with PRMan. We recognize that by the time this book is in print, there may be a number of other RenderMan-compliant renderers available to our gentle readers. We believe that the vast majority of material in this book should apply equally well to any modern and fully RenderMan-compliant renderer-and in large measure to high-quality renderers that are not RenderMan, as well.

This book is designed primarily for beginningto intermediate-level CG technical directors in the special effects industry, or those with equivalent backgrounds (such as graduate or advanced undergraduate students, or others with a strong interest in the technical aspects of computer graphics and animation). It is not assumed that you have years of production experience, or a Ph.D., or any other such exclusive prerequisites.

Preface

Nevertheless, to make this book possible, we are forced to assume that the reader brings to the table substantial background knowledge. Specifically, we assume that

■You have some background in programming, particularly in C (or a related programming language). You should be familiar with concepts such as data types and variables, conditionals and loops, subroutines and functions, parameter passing, and the basic mechanism of compiling programs. We will frequently explain concepts by merely pointing out the differences between Shading Language and C.

■You have a firm grasp of advanced high school math, including trigonometry, vector algebra, elementary functions, and analytic geometry. If you are comfortable with calculus and differential geometry, you'll be even more at home with the material.

■You are familar with the fundamental concepts of computer graphics. At the very least, the following phrases should be meaningful to you: shading and illumination models, texture mapping, geometric primitives (including polygons, patches, and NURBS), and image compositing. If you have extensively used any modeling package, animation package, or renderer, you are probably up to speed on the basic concepts of computer graphics.

In addition, there are a variety of other prerequisites that, though not strictly required, will certainly make this book much more useful to you:

■ The value of some familiarity with physics, particularly with optics and material properties, cannot be overstated. A little intuition about how light behaves in the real world can be immensely helpful in understanding how renderers work.

■We tried not to leave anything out, but this book is more a discussion of advanced topics than a tutorial or a proper reference. Therefore we do recommend that you have The RenderMan Companion (Upstill, 1990) and the official RenderMan Interface Specification (Pixar, 1989) handy. (The RI Specification is available online at www. pi xar. com.)

■It would certainly help if you had one of the several RenderMan-compliant renderers so that you can try out the examples. If you do not have PRMan, we partisanly recommend the Blue Moon Rendering Tools, which is free for noncommercial use, and can be downloaded from www. bmrt. org.

■Some knowledge of, and preferably experience with, the terminology and practice of composing and lighting of live-action and/or CG scenes is helpful. Chapter 13 provides some of this background and recommends several additional sources for this material.

A glance at the table of contents should reveal that this book is composed of very advanced material, hence the title. We have tried to provide as much context as possible by providing chapters giving introductory material on the APIs, Shading Language, and mathematical background. However, these chapters are designed more for quick reference, and while complete, they ramp up very quickly and assume substantial prior experience in certain areas. Moreover, because the material

Preface

in the book is "advanced," much of it is interrelated, without a single linear path of progressing levels of difficulty. We tried to do our best to make few assumptions in earlier chapters, building on this material for later chapters, but this was not always possible.

Well, there it is. We don't want to scare anybody off, and we have tried to make this book as self-contained as practical. But we're writing about tools actually used to make films, and they do require a lot of detailed technical knowledge.

Acknowledgments

We would like to thank all of our alpha and beta readers, Dana Batali, Rob Cook, Tom Duff, Steve May, Chris Perry, Tom Porter, Guido Quaroni, Rick Sayre, Steve Upstill, Steve Westin, and Wayne Wooten, who added so much to the clarity and technical veracity of our prose. We are also indebted to Pixar Animation Studios, Walt Disney Feature Animation, Sony Pictures Imageworks, and Warner Brothers Feature Animation for their kind permission to use images from their work.

But we are especially thankful to Ed Catmull and the whole executive team at Pixar for creating and fostering an environment in which research and development of innovative rendering techniques is actively pursued, and where, while these techniques are developed to further our proprietary animation production systems, we are still allowed to publish our ideas and discoveries and share them with the graphics community at large.

Chapter 1 is adapted from "Photosurrealism," Ninth Eurographics Workshop on Rendering, Vienna, Austria, July 1998, published by Springer-Verlag, and is used with permission. Chapter 14 is adapted from "Lighting Controls for Computer Cinematography," Journal of Graphics Tools, Vol. 2, No. 1, and is used with permission. Much of the rest of the book was adapted from sections in various Siggraph course notes, particularly Advanced RenderMan: Beyond the Companion (Siggraph, 1998), Writing RenderMan Shaders (Siggraph, 1992), and Pixel Cinematography: A Lighting Approach for Computer Graphics (Siggraph, 1996).

RenderMan is a registered trademark of Pixar. PhotoRealistic RenderMan is a trademark of Pixar. The

RenderMan Interface Specification and the documentation for The RenderMan Toolkit are copyrighted by Pixar, and excerpts are used with permission. All product names are trademarked by the companies that trademarked them.

About the Companion Web Site

Readers are encouraged to visit the companion Web site to this book at http: //www . mkp. com/renderman. At the Web site you will find all of the code in this book, both library routines and full shaders. No need to type in any program listings-plus, the

Preface

Web site's files will include any extensions and corrections that might be made after the book is printed. In addition, the Web site contains other examples, pointers to RenderMan-related URLs, and other resources, as well as an area where we can post additional examples and shaders, written and contributed by users, that are based on the material in the book.

Photosurrealism

Over the last 25 years, a revolution has been taking place in the motion picture industry. It started slowly at first, faltered at times, but has been growing and building inexorably. What once was a mere curiosity used to obscure ends is now the basic paradigm of film production. Movies have gone digital.

There are many aspects to movie making, many processes and techniques where digital technology can and has made a difference in the way that movies are made-everything from text editors customized for script writing to digital soundtracks on the film (and eventually digital distribution replacing film itself). But perhaps none have been so revolutionary as the use of computer graphics to create imagery for motion pictures.

The first use of computer graphics in a feature film was in 1976 when Futureworld included a computer-animated simulation of a human hand and face. These animations, created by then graduate students Ed Catmull and Fred Parke at the University of Utah, started the

1 Photosurrealism

computer graphics industry down the road that eventually led to the nearly ubiquitous use of computer graphics special effects in feature films.

The road was long, however. The first two films to make significant investments in computer-generated imagery (CGI), Disney's Tron and Universal's The Last Starfighter, were commercial flops. These failures made directors wary. Generally, CGI was limited to imagery that was clearly supposed to be from a computer (fancy displays on computer monitors, for example). It wasn't until 1990 that CGI scored its first real success:

won the Academy Award for Best Visual Effects, partially on the strength of the photorealistic CGI effects produced by Industrial Light and Magic.

Since 1990, everything has changed. Computer graphics has been used to create significant special effects for every film that has won the Academy Award for Best Visual Effects. Many Scientific and Technical Academy Awards have been awarded to the authors of computer graphics software, recognizing their contributions to the film industry. Presently, nearly all films that have any visual effects at all will have at least some effects that are computer generated. Most movie special effects houses in the United States have largely switched over from "practical" effects to CGI-based ones. Moreover, those visual effects are not limited to spaceships and monsters. Main characters with significant screen time can now be created with computer animation, as was seen in

Phantom Menace.

Or we can dispense with live-action footage entirely. The first full-length completely computer-generated feature film, Pixar's Toy Story, was released in 1995 and was a huge commercial success. As this book goes to press in 1999, two more successful all-CG films have already been released, and we are aware of at least 10 more that are in some stage of production.

Computer graphics has carried the day.

1.1Making Movies

For more than 20 years, one of the stated goals of the computer graphics research community has been to solve the problem of making truly photorealistic images. We have strived to make the CG image, as much as possible, look exactly like a photograph of the scene, were that virtual scene to actually exist. Much of this work was done very specifically so that CGI could be used in films. Solving such problems as accurate light reflection models, motion blur, depth of field, and the handling of massive visual complexity was motivated by the very demanding requirements of the film industry.

Movies, however, are illusions. They show us a world that does not exist, and use that world to tell a story. The goal of the filmmaker is to draw the audience into this world and to get the audience to suspend its disbelief and watch the story unfold. Every element of the movie-the dialog, costumes and makeup, sets, lighting, sound and visual effects, music, and so on-is there to support the story and must help

Altered Reality

lead the viewer from one story point to the next. In order for the audience to understand and believe the movie's world, it clearly must be realistic. It cannot have arbitrary, nonsensical rules, or else it will jar and confuse the viewers and make them drop out of the experience. However, movie realism and physical realism are two different things. In a movie, it is realistic for a 300-foot-tall radioactive hermaphroditic lizard to destroy New York City, as long as its skin looks like lizard skin we are familiar with. Some things we are willing to suspend our disbelief for, and some things we are not!

There are two people who are primarily responsible for determining what the audience sees when they watch a movie. The director is responsible for telling the story. He determines what message (or "story point") the audience should receive from every scene and every moment of the film. The cinematographer (cinema photographer) is responsible for ensuring that the photography of the film clearly portrays that message.

Over time, filmmakers have developed a visual language that allows them to express their stories unambiguously. This language helps the filmmaker to manipulate the imagery of the film in subtle but important ways to focus the audience's attention on the story points. Distracting or confusing details, no matter how realistic, are summarily removed. Accenting and focusing details, even if physically unrealistic, are added in order to subtly but powerfully keep the audience's attention focused on what the director wants them to watch. The job and the art of the cinematographer is to arrange that this is true in every shot of the film. The result is that live-action footage does not look like real-life photography. It is distilled, focused, accented, bolder, and larger than life. In short, film images manipulate reality so that it better serves the story.

Our computer graphics images must, therefore, also do so. Perhaps even more than other parts of the filmmaking process, CGI special effects are there specifically to make a story point. The action is staged, lit, and timed so that the audience is guaranteed to see exactly what the director wants them to see. When a CG special effect is added to a shot, the perceived realism of the effect is more influenced by how well it blends with the existing live-action footage than by the photorealism of the element itself. What the director really wants are images that match as closely as possible the other not-quite-real images in the film. They are based in realism, but they bend and shape reality to the will of the director. Computer graphics imagery must be photosurrealistic.

1.2 Altered Reality

In a CG production studio, the job of the cinematographer is often divided among several CG artists whose jobs include geometric modeling, camera placement, material modeling, lighting, and renderer control. These digital cinematographers are

1 Photosurrealism

collectively known as technical directors. They are the users of image synthesis software.

Computer graphics gives technical directors a whole host of new tools, removes restrictions, and provides an interesting and growing bag of tricks with which to manipulate reality. Jim Blinn once called these tricks "The Ancient Chinese Art of Chi-Ting" (Blinn, 1985). Technical directors now call them "getting work done."

1.2.1Nonphysical lighting

One of the best cheats is to alter the physics of light. Chapter 14 discusses a light model used in computer graphics cinematography that was specifically designed for this type of flexibility. This work is based on the observation that movie viewers generally do not know where the lights are located in a scene, and even if they do, their limited 2D view does not allow them to reason accurately about light paths. Therefore, there is no requirement that the light paths behave realistically in order to ensure that an image is believable. This situation is exploited by the live-action cinematographer in the placement of special-purpose lights, filters, bounce cards, and other tricks to illuminate objects unevenly, kill shadows, create additional specular highlights, and otherwise fudge the lighting for artistic purposes. Everyone who has ever seen a movie set knows that even in the middle of the day, every shoot has a large array of artificial lights that are used for a variety of such purposes. However, the live-action cinematographer, despite his best tricks, cannot alter the physics of the light sources he is using. Computer graphics can do better. Creating objects that cast no shadow, having lights illuminate some objects and not others, and putting invisible light sources into the scene near objects are all de rigueur parlor tricks. Things get more interesting when even more liberties are taken. Consider, for example,

*a light that has negative intensity, and thus dims the scene

*a light that casts light brightly into its illuminated regions but into its "shad owed" regions casts a dim, blue-tinted light

*a spotlight that has at its source not simply 2D barndoors to control its shape, but instead has full 3D volumetric control over the areas it illuminates and doesn't illuminate, and at what intensities

*a light that contributes only to the diffuse component of a surface's bidirectional reflection distribution function (BRDF), or only to the specular component

*a light that has independent control over the direction of the outgoing beam, the direction toward which the specular highlight occurs, and the direction that shadows cast from the light should fall

1.2.2 Nonphysical optics

Another interesting cheat is to alter the physics of optics, or more generally all reflection and refraction, to create images with the appropriate emphasis and impact.

Much of the trickery involving manipulating optical paths is descended, at least in part, from limitations of early rendering systems and the techniques that were invented to overcome these limitations. For example, for over 20 years, most CGI was created without the benefit of ray tracing to determine the specular interreflection of mirrored and transparent objects. Even today, ray tracing is generally considered by the production community to be computationally too expensive to use except in particular extreme circumstances. However, mirrored and transparent surfaces still exist, and ways must be found to render them.

As before, the limited knowledge of the viewer comes to our rescue. Because the viewer cannot estimate accurately the true paths of interreflection in the scene, approximations generally suffice. In place of exact reflection calculations, texture maps are used: environment maps are accurate only from a single point in space (the center of their projection) but work on all surface types and in all directions; planar reflection maps are accurate everywhere on a planar surface but only when viewed from a particular camera location. Use of environment maps "near" the projection point and use of planar reflection maps on "almost" flat objects lead