Defining the Next Generation of Real-Time Characters
Posted March 29 10:48 PM by Kelly Murdock
Paul Steed presented a session on Friday titled, “Defining the Next Generation of Real-Time Characters.” In this session, Paul used images of characters taken from many current games as a way to highlight examples of good and poor character design. He began the session with a statement that, “a good character isn’t just more polygons, it’s a better character.”
He then highlighted four principles that are used to create distinguished characters. These principles include:
Higher Visual Resolution
Distinct Material Definition
Better Deformation While Animating
Higher Visual Resolution In the area of higher visual resolution, several specific elements combine to produce characters that look better including:
Product Design: Whether sketched or painted, the original design needs to look great.
More Polygons: The overall detail of the character needs to be enough so the character looks smooth. This also applies to the environment, which also needs to be at a same quality level as the characters.
Normal Maps: Future next-gen characters can possibly include animated normal maps. A good example of effective normal maps is Project Offset.
Bigger Texture Maps: Larger texture maps allow more details to be added to characters.
More Bones: More bones yields better deformation and a wider range of motion.
More Effects: Better effects make the character look better.
More Characters: Crowds of characters can be impressive.
LOD: LOD can even be applied to bones.
HDR: HDR images helps the environment feel more cinematic, and they make the character stand out more.
Paul concludes this section with a plea, “there is no reason to have low-poly characters in next-gen games.”
Distinct Material Definition Another key way to differentiate your characters is with distinct material definitions. This includes the following:
Specular Maps: Even if a specular map includes nothing but noise, it is better than nothing.
Damage States: The characters in the latest Fight Night game are great examples of how damage states can be used.
Subsurface Scattering Effect: This effect can be used to show light through the back of the ear. It is also good to add an edge lighting effect to the scene.
Dynamic Reflection Maps: These maps can add a lot to the visual appeal of a scene.
Texture Randomization: This procedure can add diversity to the scene.
Same Texture-Different Versions: Reusing textures in unique ways.
So many of the character materials in the past have appeared plastic. We need to ask ourselves, “Are we maxing out the technology?” There are new techniques and effects that we need to begin taking advantage of, like steep parallax maps.
Better Deformation While Animating Characters that move realistically, look realistic. There are several elements that can help characters deform better during animation sequences including:
Modelers Need to Animate: A modeler that animates has a better idea of how to build characters that deform well. Modelers should also actively study anatomy.
Better Rigs: Fighting games have traditionally always had the best rigs because they do the most moves, but next-gen games should all have this level of character motion.
Animation Testing: By testing how models animate, modelers can learn a lot on how to make them better.
Accurate Weighting: Good weighting needs to be included on all characters in the scene including the supporting cast and not just the main hero.
Morph Targets with Skeletal: Although morph targets have traditionally only been used for facial animation, full body morph targets can now be used for next-gen games.
Motion Blur: This effect can make a scene feel more real. It also helps hide poor deformation.
Ragdoll Physics: Using ragdoll physics, you can catch many motions that you normally miss. For example, when you walk down stairs, you typically tilt your head to look at the steps. This doesn’t occur when walking up stairs, but only when walking down.
Procedural Animation: This is an interesting area that will become more important in next-gen games.
Real-Time Tessellation: This effect is the no. 1 request of animators. It enables them to use a low-end rig and tessellate the model in real-time.
Advanced Characteristics In addition to the standard character features like modeling and materials, many advanced characteristics can be used to give characters the extra something they need including:
AI: In next-gen consoles, a separate GPU can be dedicated to calculate AI.
Realistic Emotion: Next-gen characters should be able to create real emotion rather than simply making a face.
Accurate Self-Shadowing: This should allow all objects to be shadowed and to cast shadows on characters in the game without requiring that the objects receiving shadows are closed.
Dynamic Normal Maps: This next-gen effect that could be used to yield some interesting results.
Scale Randomization: Next-gen games should feature crowds where everyone isn’t the same height.
Procedural Hair Simulation: Using spline-based hair, you can take hair simulations to the next level.
Procedural Cloth Simulation: The same applied to cloth as to hair, better realism.
Whole Body Morphing: This would allow you to simulate the effect of veins popping out when the arm is held down due to blood rushing into the arm.
Specifically, Paul highlighted the Fight Night game as a good example of the type of characters that need to appear in next-gen titles. With its attention to detail, textures and realistic body movements that take advantage of the new technology, this game is a prime example of what is possible.
Final Thoughts Paul concluded the presentation with several suggestions for character modelers. He encouraged all artists to get more technical. Take time to study the various platforms, look through the art assets included in the SDK, and talk with the programmers. Get involved in the pipeline, experiment and learn. Through these efforts, your skills will increase and the resulting characters will be better.