Continuing our excerpts from the Inspired 3D series, Tom Capizzi presents an in-depth character construction tutorial in three parts.
This excerpt is the next in a number of adaptations from the new Inspired 3D series published by Premier Press. Comprised of four titles and edited by Kyle Clark and Michael Ford, these books are designed to provide animators and curious moviegoers with tips and tricks from Hollywood veterans. The following is excerpted from Modeling & Texture Mapping.
In this chapter the focus of the discussion will be the character and not the development of the environment. This is not to say that the environment is not important, but the process will be easier to detail if the focus is limited to one character.
James Van Der Kyle created the storyboards in Chapter 12 to produce an animated short film. This short animation was created for the tutorials in this book and other books in the Inspired series. The design for this character shown in these storyboards was not the original design. Figures 1 and 2 show the original designs.
[Figures 1 & 2] The concept and the model for the character that were originally created for the book series.
Process for Building the Character
The sketch shown in Figure 1 was created by James van der Keyl. This sketch has an appeal that was exceedingly difficult to capture in 3D. The model shown in Figure 2 was created by Daniel Dawson. This model has a certain appeal, but it does not quite have the appeal of the sketch. In truth, although the design did capture the original better than the other models produced during the design process, this model still had many problems.
Aesthetically, the model lacked a certain amount of realism that the sketch conveys. The details and proportions are simplified to create a cartoonish look. Technically, there were problems with the construction that would make the model hard to animate, specifically, the mouth and eyes are not in neutral positions. When a characters face is modeled, the eyes and mouth are usually modeled in a relaxed position. This position appears as a 50% open and 50% closed position. This makes the construction of face targets easier; the mouth is already halfway open, so it is simple to open it, or it can be closed just as easily.
In defense of the modeler who created this model, Daniel Dawson, the problems associated with this model were directly attributed to the direction he received, and this model represents a first pass at the final model. Daniel was given no opportunities to finesse this model.
This model was chosen to proceed with the series illustration. Now that this decision was out of the way, I still needed to create a model for a tutorial. The model shown in Figure 2 used a polygonal construction process, and a subdivision surface smoothing pass to finish the model. Although this is a valid way to create a character model, I wanted to write a tutorial that used NURBS surfaces as a base for the subdivision polygons.
The model that eventually became the one used for the book was originally constructed as a model that used patch modeled surfaces as a basis for the final construction.
[Figures 3, 4 & 5] These diagrams illustrate the design direction taken for the final character.
Final Modeling Process
The character in my original sketches (Figure 3) did not have the edge I was looking for, so I made the jaw line more natural, and the eyes were brought forward more (Figure 4). Because the design I was coming up with only had to make me happy, I did not concern myself with the cute aspect of the character design. The character I was trying to build looked like the technical directors I know edgy, burned out, but overall fairly nice people.
Upon viewing the first rendered image (Figure 5), I decided that this character had several advantages over the character that was originally chosen. This character had a certain appeal, but the appeal was more oriented toward adults instead of children. This character was more edgy.
The final modeling process this completed character started with a NURBS patch model. The NURBS surfaces were converted to low- resolution polygons soon after the NURBS surfaces were laid out, and finally were rendered using subdivision techniques. This workflow allowed for the quick and accurate fabrication of a model that could be edited efficiently and repeatedly.
This process is efficient for several other reasons that extend beyond the task of modeling.
- 1. The UVs created on the low-resolution model are transferred to the subdivided model during the subdivision process, making texture mapping very fast.
2. Animation setup is done on the low-resolution model. The controls, vertex weighting and bindings created in animation setup are all maintained during the subdivision process.
3. Animation can be done on the low-resolution model, which allows for the fast manipulation of the low-resolution deformable model. The intersections and excessive wrinkling that can occur during the animation process can be prevented by increasing the resolution of the low-resolution model after animating it, and previewing the results. This gives animators instant feedback for their animation.
4. The actual subdivision can be a control that is set right before rendering the model. This makes the entire pipeline faster.
During the process that was used to get this character from concept to cover art, the ability to change the model, sometimes dramatically, was essential. This workflow put the right amount of detail in the right location. This allowed the model to change many times without distorting into a ghastly creature, like the previous attempts. This also allowed the changes to happen without changing the original design intent while the character was in development.
NURBS Patch Modeling
The process used to create the final model started with a NURBS model based loosely on the original sketches (Figure 3). Throughout the course of the project, this character took on a larger role, and the NURBS patch modeling was never completed to a finished level. The models shown in Figures 4 and 5 are fairly preliminary and are useful in looking at the way the character feels. These models have a long way to go.
The construction of the patch model started with rough splines sketched in the modeling window. The rough splines were filled in with loose patches until there was a low-resolution network of surfaces.
[Figures 6 & 7] The shapes of the muscles of the face correspond to the surface layout used in the NURBS.
[Figures 8 & 9] The patch surface layout of the ear.
[Figures 10 & 11] The patch model has several advantages to the single-surface modeling paradigm.
In this example, the correlation between the muscles of the human face (Figure 6) and the patch layout (Figure 7) can be seen. There is a radial layout of surfaces around the eyes, ears and mouth, and the surfaces of the cheek and eyebrow follow the contour of the skin.
The original NURBS ear is shown in Figure 8. The surface layout of the ear is similar to the surfacing layout used at the mouth and eye. Using a single surface to describe the center of the ear and having the detailed areas of the ear radiate out from the center are efficient ways to get the required detail of the ear and have this complex area transition to the smooth area of the cheek. Figure 9 shows a diagram of the surfaces.
Why a NURBS Patch Model First?
In earlier chapters, the subject of creating a model that has parameterization that follows the flow of the geometry is mentioned. Modelers can use other modeling paradigms for building a characters face, but no method handles the issues related to building a characters head as well as the patch modeling method. Other modeling techniques, such as the single-surface head, do not cover all of the issues related to surface layout as well as the patch modeling method.
First of all, the flow of the geometry does not match the musculature beneath the skin as well. And the single- surface head (Figure 10) has to overcompensate for the detail in the nose and eyes by adding isoparms. The patch model (Figure 11) can use less geometry and include just as much, if not more, detail. Most important, the extreme stretching that occurs at the back of the face and around the chin on the single surface head does not occur at all on the patch model.
Overall, a patch model has a more even distribution of geometry in the proper location, which makes this a great starting point for building the rest of the model.
[Figures 12 & 13] This shows how the NURBS surfaces convert to polygons (left) to create the framework for the polygonal model. The model (right) is shown after some preliminary editing.
Convert to Polygons
The NURBS surfaces are converted to polygonal meshes. The settings used to convert the NURBS surfaces to polygonal meshes vary depending on the density of polygonal mesh that the model needs to be. In Figure 12, the correlation between the conversion of the polygonal mesh and the NURBS surfaces can easily be seen. The NURBS surfaces and the polygonal mesh are superimposed on top of each other.
Initially, the distribution of the polygons is too heavy in some places and not dense enough in other places. The polygonal model has to be adjusted by hand in order to get the detail distributed in the right places. The polygonal model shown in Figure 13 was converted directly from the NURBS surfaces. This model has to be edited before it can be used.
When a NURBS mesh is converted to polygons, the problem of badly distributed geometry must be addressed before the model can be used. The first problem that is normally taken care of is the problem of too much detail in the wrong areas. Some places on the head maintain a generally smooth topology. Places like this are the top of the head, the back of the head and the cheeks.
During the creation of a NURBS patch model, some additional isoparms are added to certain areas just to get the parameterization of these areas to line up with other parts of the model. For example, the top of the head will often get additional geometry that has been generated in the nose area because the top of the head rests directly in line with the nose. This is one of the complications that must be dealt with when working with NURBS surfaces.
Now that the model has been converted to polygons, this is no longer an issue. Now the high-resolution geometry on the top of the head can be decreased. This can be done by eliminating edges, snapping vertices to a common point (creating polygons with zero surface area) or by deleting vertices. Decreasing the resolution achieves two goals:
- 1. This reduces the geometric complexity of the area, which helps reduce the possibility of creating wrinkles and unwanted flashing in that area.
2. By reducing the number of polygons, the area will appear smoother when subdivision surfaces are applied to those polygons. (Figure 14.)
Two guidelines should be adhered to when optimizing polygonal models:
[Figures 15 & 16] When areas that have different geometric complexity run into each other, a transition will occur. These areas will require triangles to get the transition to work.
- 1. Try to avoid triangles. Sometimes reducing geometric complexity in a particular region will cause a transition to another, more detailed area. In cases like this, the low-resolution rows of polygons will run into higher-resolution rows of polygons. The transition has to be a series of triangles that get the two areas to smoothly flow into each other (Figures 15 and 16). This is an unavoidable circumstance, unless the polygonal model is not optimized at all after converting the model from NURBS surfaces. Certain problems will arise, however, if the model has too many triangles. Models that are all triangles are almost unusable for subdivision modeling techniques. Although triangles are sometimes a necessary evil, they should be avoided.
- 2. Try to maintain an even distribution of geometry if possible. There will be areas that will require fine detail, but the quadrangles on the cheek should maintain relatively the same size as the quadrangles on the top of the head. This will allow the textures to map more evenly to the geometry and will facilitate the use of subdivision modeling much easier than if there is a dramatic change in size between the polygons in different areas of the model.
To learn more about character modeling and other topics of interest to animators, check out Inspired 3D Modeling and Texture Mapping by Tom Capizzi; series edited by Kyle Clark and Michael Ford: Premier Press, 2002. 266 pages with illustrations. ISBN 1-931841-49-7. ($59.99) Read more about all four titles in the Inspired series. Read Part 2 and Part 3 on VFXWorl.
Tom Capizzi (left), Kyle Clark (center) and Mike Ford (right).
Tom Capizzi is a technical director at Rhythm & Hues Studios. He has teaching experience at such respected schools as Center for Creative Studies in Detroit, Academy of Art in San Francisco and Art Center College of Design in Pasadena. He has been in film production in L.A. as a modeling and lighting technical director on many feature productions, including Dr. Doolittle 2, The Flintstones: Viva Rock Vegas, Stuart Little, Mystery Men, Babe 2: Pig in the City and Mouse Hunt.
Series editor Kyle Clark is a lead animator at Microsofts Digital Anvil Studios and co-founder of Animation Foundation. He majored in film, video and computer animation at USC and has since worked on a number of feature, commercial and game projects. He has also taught at various schools, including San Francisco Academy of Art College, San Francisco State University, UCLA School of Design and Texas A&M University.
Michael Ford, series editor, is a senior technical animator at Sony Pictures Imageworks and co-founder of Animation Foundation. A graduate of UCLAs School of Design, he has since worked on numerous feature and commercial projects at ILM, Centropolis FX and Digital Magic. He has lectured at the UCLA School of Design, USC, DeAnza College and San Francisco Academy of Art College.