[Silent Hill character designer, Fatale contributor and CG artist Takayoshi Sato, currently art director at Virtual Heroes, examines the art of creating emotionally involving computer-generated characters in this in-depth feature, originally published in Game Developer magazine.]
Technology has greatly improved since the first 3D consoles appeared. Back then, as the more veteran artists may recall, one building was less than 50 triangles, and a character was less than 500. The standard for a texture was 16 colors at 128 X 128 resolution. It was extremely challenging to render anything organic. We needed to work within severe limitations, and players had to accept pixelated textures all over the place.
Now we've got higher console and PC specs, and the luxury of technologies such as normal, height, and specular maps, ambient occlusion, sub surface scattering, and lots of real-time rendering features that can compete with software rendering.
I recall around 2005, close to the release of the Xbox 360, we started hearing about the Uncanny Valley, and the term has only gained traction since then. This means despite the fact that rendering quality has improved tremendously and characters are rendered with high detail, they still lack emotion, and our extremely detailed models are acting like marionettes. The better rendered the model is, the weirder it tends to look.
3D scanning and motion capture cannot be completely trusted either. Motion capture has its own odd look, and 3D scanned heads often don't look anything like the original person.
Partially this is because some sort of distortion occurs in the process, but more than that, we don't get the feeling of a person from that frozen still head, like we do not get that feeling of humanity by looking at decapitated head separate from its body.
Once I was involved with a very complex experimental scanning process, scanning a 3D shape at 60 frames per second, while recording texture and voice at the same time. It should have captured all human essence in one shot. It should have looked very lively. But it did not.
These soulless models playing out actions before us leave us with an empty feeling. Characters need to be emotionally engaging, touching, and must dominate the scenes they're in with their presence.
This cannot be done with any type of recording. Character creation spans several areas outside of the character department (in a typical game production) -- in other words, a solid character cannot be established without many disciplines working together as a team, managed by a strong vision. It needs to be elaborately planned, because creating attractive characters is essentially the same as creating good scenes or stories.
In this article I will try to break down the most important elements involved in the creation of an emotionally involving character.
A good story consists of several key moments connected together, and characters exist as a vehicle with which to navigate these key moments. So the first thing to do is create and understand your key moments.
There are two prominent shots that we need to pay extra attention to in regard to character. The first is the introduction shot. This is literally an introduction to the character, and provides a first impression for players.
Second is the reaction shot, focusing on a character after a given important event happens. It shows the character's face, telling the audience what kind of incident it was for the character, which helps to define their position within the story. Usually in linear media the scene ends with this reaction shot, and it remains in the audience's mind as a major story impression, whether they are conscious of it or not.
You cannot really do anything without understanding and determining the mood. A simple character description is just not good enough. A figure could look totally different depending on the mood surrounding it.
For instance, a smile that looks like a horrifying grin in Rembrandt lighting might appear innocent with a Renoir lighting scheme. Or if the subject is a monster, with a Brueghel-like background it may look like a legendary story, but would appear to be more of a surreal fantasy in Odilon Redon's style.
Core elements such as story, theme, and philosophical message determine the mood of the target scene. On a more subliminal, but no less important level, lighting and composition are main contributors that define this mood visually. Characters can look totally different under different a lighting scheme, lens, or scene composition.
Important shots need to be designed with those elements in mind, especially in cut scenes or establishing/reaction shots. Ultimately we are aiming for that perfect shot, and character development ideally starts with the scene test bed with the lighting and camera prepared.
Figure 1: Angela from Silent Hill 2.
In the image of Angela from Silent Hill 2 (see Figure 1) there is not much obvious facial expression or any emotional depiction on the character, but you can tell by the subtle things like lighting and the camera setting that there is enough room behind her for somebody to attack, and there is someone (the player in this case) in the scene -- and she knows, but does not care much.
There is very stable screen construction following her bodylines, yet the banked camera conveys a somewhat unstable moment. She seems to be indulging in her own moment defenselessly. Her eyes tell us there is nothing important in the direction she is looking other than the knife, but her eyes don't focus on it. This image is giving us a lot of information without using the typically overt emotional tools seen in games.