3-D animator Bill Fleming details the development of technologies for animating fur and hair in computer animation.
Is your 3-D character having a bad hair day? There seems to be an epidemic of bad hair on 3-D characters. Of course, it's not the fault of the artists as much as it's the technology. While 3-D programs have come a long way, they have yet to really perfect hair effects. Hair is the defining element of many creatures. Unfortunately, for the 3-D artist, this presents a real challenge since hair and fur in reality are comprised of tens of thousands of tiny fibers.
Therefore, to make hair appear realistic, you literally need to have the same number of fibers on your 3-D character. This can be a real challenge, particularly if you plan to model the hair by hand. Another complication of hair is the fact that it tends to vary in length, density, color and thickness. It can be curly, kinky, straight, wavy or even braided. Plus, there's the hairstyle to consider. Not too many characters look appealing with a "bowl" cut. As you can see, there are many factors to consider in creating hair for your 3-D character. So, how do you go about creating the natural hair look for your characters? Well, let's take a look at the growth of 3-D hair over the years and the many techniques that have evolved along the way. The Evolution of 3-D Hair Ten years ago 3-D hair was nothing more than a plastic shell. This is the original 3-D hair, which was rather stylized and cartoony. The hair was created with simple geometry that formed the volume of the hair as shown in Figure 1.1.
Typically, bump maps were used to simulate strands of hair but they were often no more than straight lines, which left the hair looking plastic and plain. This, of course, is a nice look for cartoon characters but there are times when a more natural effect is needed, such as in those high budget Hollywood effects films. Another problem was that the hair couldn't be animated. It just sat there, solid as a rock and lifeless. The good news about this style though was that every 3-D program was capable of creating it. There were some very innovative artists who styled the mesh to create simulated hair styles such as the popular Japanese anime look, which was definitely a good look, but still wasn't animatable. This primitive hairstyle was the only option for many years until 1989 when we made the next leap in technology by developing the Clip Map.
Clip Map Hair
A clip map is a black and white image that's used to clip portions of a model. The white areas of the clip map represent the part of the model that will be clipped, or cut off. Figure 1.2 shows a clip map and the accompanying color map for hair.
Using clip maps, we could simulate hair strands by clipping fine lines in a hair mesh. This, coupled with a nice hair image map, made it possible to create relatively realistic hair in still images, like the one shown in Figure 1.3.
Hash's Animation Master, LightWave, 3D Studio MAX, Image, Softimage, and several others. Clip map hair looked much better than the classic stylized hair but the hair was flat, lacking any depth. It also presented the same problems in regard to animation. The hair still hung lifeless. Well, the animation problem was temporarily solved by the introduction of bones in 1990.
Bones allowed us manually to move the hair mesh to produce movement. Of course, the only shortcoming was that the hair moved in large clumps rather than strands waving in the breeze. Basically, we could make a polygon ponytail wiggle in the wind but it had to move as a single unit, rather than as many hair strands whipping around. The next step in hair evolution was the introduction of geometry replication.
Geometry replication utilities made it possible to build a single strand of hair and replicate it over the surface of the mesh. This was a significant improvement in 3-D hair styling but the hair tended to appear rigid and moved without the implications of physics as we can see in Figure 1.4.
Generally, it looked artificial because the hair was typically sticking straight out from the head like a ball of spikes. Many programs support geometry hair such as LightWave, 3D Studio MAX, Softimage and Alias PowerAnimator. This technique evolved to the point where these replicated strands could be styled with random effects such as jitter, contour, curl and kink. Contour was the most significant addition since it allowed us to mold the hair strands to the shape of the head. These new tools were provided in the form of plug-ins such as MetroGrafx's Fiber Factory for LightWave, AFX's Furrific for RayDream Studio and Peter Watje's Scatter for 3D Studio MAX. Of course, we still had little control over the actual hairstyle so a salon cut was out of the question but we could achieve some great effects as seen in Figure 1.5.
Soon technology advanced and we were able to perform styling on the hair strands, but then we ran into the problem of exceedingly high polygon counts. Ten thousand hair strands can really rack up the polygons. There was also the problem of animating these hairs. While morph targets enabled us to move the hair, it was all moved in one big motion rather than randomly shifting like hair does in reality. One solution to this problem was to use a small, animated fractal displacement map to randomly shift the hair. This actually worked very well to simulate the effects of wind on the hair but offered no solution for simulating the movement of hair when the body of the character moved.
To solve this problem, animators used "control" objects to animate the hair. Basically, several control hairs were created, which affect the movement of many other hairs through the use of instancing. Instancing allows one to make reference copies of objects. One has a single object that controls the movement of several others. With this technique, one could have a group of hairs on the top of the head controlled by a reference object and another group on the side of the head controlled by yet another reference object. This is a bit of a tedious way to animate but the result is more controlled than displacement mapping.
Another solution to animating geometry hair is the use of soft body dynamics, which allows one to move the hair naturally with collision detection to prevent the hair from penetrating the head and body. Both Maya and 3D Studio MAX have soft-body dynamics systems. The Maya soft-body system is available as an FX upgrade while the MAX soft-body system is a plug-in called Hypermatter, developed by Second Nature Systems. The only shortcoming of both the soft-body and "control object" techniques is that they tend to affect groups of hairs rather than individual strands. Of course, the effect is certainly better than the hair being motionless.
One of the main drawbacks of geometry hair is the hit on render times. Since the hair is made of actual geometry, it can really impact the render time, particularly if one is animating the hair. To resolve this problem environmental hair effects were created.
Environmental hair is considerably faster to animate since actual geometry isn't created. Both Softimage and 3D Studio MAX introduced environmental hair options in 1997.
Environmental hair is available in 3D Studio MAX through the Shag: Fur plug-in created by Digimation. Shag: Fur uses image maps to determine exactly where the fur is applied, the density, color, thickness, direction, leaning and bend of the hairs. What's more, separate texture maps can be used for most of these options to provide complete control. For example, a texture map of a tiger skin can be used for fur color so that the fur hairs derive their color from the map image, while a separate map can be used to control where the hairs are thick and thin. Almost all of these parameters are animatable, so subtle motion, growing and color changes are all possible. Of course, Shag: Fur does not do dynamics or any type of automatic movement but it does create some very convincing hair effects as shown in Figure 1.6.
The environmental hair solution for Softimage is included in Mental Ray. Mental Ray is similar to Shag: Fur but it has one shortcoming in that it doesn't calculate the strands on the back of the object. Therefore, when your character/creature turns, the strands on the front of the body will disappear.
The best solution for creating animated hair is the use of particle systems.
Particle hair is the most commonly used method for creating the Hollywood hair effects like the ones seen on the lion's mane in Jumanji and the werewolf in American Werewolf in Paris. Unfortunately, particle-based hair systems are typically custom developed for specific film and broadcast projects.
Particle hair systems work by emitting particles from the skin until they reach a death point which is the desired length. Then, using collision detection and physics the system controls the hair so it behaves naturally by moving and reacting with the character. For example, in the scene in American Werewolf in Paris where the werewolf exits the fountain and shakes the water off, the particle hair is reacting to the physics of the werewolf's movement.
There is one commercial particle hair system currently available from Alias/Wavefront called Compuhair. Compuhair calculates hair as soft volumetric tubes with a fuzzy volume, something like a cloud with soft edges. Compuhair allows one to control the hair bounce through the use of control spheres. There are several particle hair systems in development for other products such as Nordisk Film's Fur Designer for Softimage, which renders hair on a separate pass and then composites it back on the scene.
As you can see, there are a number of options for creating hair on your characters. It really depends on the volume of hair and the animation you'll be doing. Oh yes, and how deep your pockets are. Computer graphics hair effects have come a long way in a short time. Within a year the now unattainable particle hair systems will be available on many of the mid-range 3-D programs. Until then, try experimenting with the many options that are currently available, you'll be surprised at the effects you can achieve.
Bill Fleming is president of Komodo Studio, a 3-D studio specializing in photorealism. He is the author of many 3-D books, including the 3D Photorealism Toolkit, published by John Wiley & Sons. He also serves as editor in chief of Serious 3D magazine, a 3-D magazine featuring nothing but intermediate/advanced tutorials for artists interested in taking their 3-D graphics to the next level. To find out more about Serious 3D, visit www.serious3d.com.
ASIFA And Festivals: A Changing RelationshipPrevious Post