Digital Art Masters: V7 - Sample Making Of 'Qev'
Qev was conceived as an internal R and D project at our character art studio Slide. He was primarily the result of an exercise in creating a face that was believable, yet entirely imagined. However, as we always try to squeeze the most out of our internal projects, one of our other main aims with Qev was to produce a model that would look great at both mid-distance and in extreme close-ups. This is something often requested for cinematic jobs where shots are always subject to unexpected, last minute changes.
As art director/artist on the project (working with the co-founder of Slide, Robin Deitch, who helped formulate the technical approach) I knew from the start that I wanted Qev to be an anti-hero from UK youth culture with a sci-fi twist. I wanted him to display all the stereotypical traits affiliated with the strata of UK culture known as "chavs", but also wanted to subvert these with a few elements that would make the character sympathetic. I decided that these elements would primarily be conveyed through the design of his face, which we wanted to be realistic, but entirely imagined so as to give us maximum creative control over his appearance.
The wish to create a realistic, yet characterful, face was both an ongoing personal goal and something Slide as a studio had seen a greater demand for in the games industry as we moved towards the end of the current gen cycle and into the next generation. Generally in this console cycle detailed human characters have commonly been developed in one of two ways: illustrative and hyper-real (usually based on concept art) or actor photo/scan based and "photoreal". Each of these methods has strengths we wanted to tap into and weaknesses we wanted to try and avoid. In the case of the more illustrative approach, the strength lies in the ability to convey a lot of personality and have full creative control over the look of the character. However, going down a strictly illustrative approach means characters can end up looking archetypal and lack subtlety and depth. On the other hand, in the case of an actor/scan-based approach you can achieve wonderfully subtle results, but are restricted by the look of the actors you cast. This, in turn, limits your creative choices and you run the risk of ending up with a "death mask" quality once the character is in the game.
So our goal was to develop our skills and techniques for the purposes of realistic human facial design and harness the strengths of both these methods. Qev was our first full production character created using this process; an approach we have since fine-tuned and successfully used across multiple AAA projects coming out over the next year or two. The process relies on observation and artistic judgment, so it is not something that can be conveyed in a tutorial. At a high level though the key to it lies in learning the bounds, limitations and constants of the human face in real-world terms, and developing an instinct for how to bend these rules to achieve a specific result without destroying the illusion of realism.
A key to our approach, which helps the process, is to get broad color information on a head model as early as possible and regularly create renders under varying lighting conditions using appropriate shaders as we sculpt (Fig.01 - 02). This helps to catch areas and features that may have become too extreme during the sculpt or that just need a little more "TLC" to really sell them.
This process took several days on Qev as it was my first attempt at formulating my approach into a repeatable, production-ready technique. Once this was done we had a face and likeness around which we could start to build the character.
For his costume design we decided to develop the idea of sports leisure wear by clothing him in a tracksuit and incorporating a cybernetic exoskeleton used to augment running (presumably from the police while carrying a recently "discovered" HD TV). This contrast of cloth and hard surface structures would allow us to test out a pipeline for multiple high resolution texture tiles across both displaced and subdivided geometry, as each have their own specific requirements and foibles (Fig.03).
When sensible UVs had been assigned and we could apply a measured checkerboard map, it became clear that we'd need to spread the UVs over a number of pages to maintain the desired texel resolution, whilst keeping individual textures at a workable size in Photoshop.
Materials in 3ds Max can usually only take a single bitmap as a texture input, but we wanted to avoid duplicating our shader networks to handle the multiple source textures. Consequently, we laid out the UVs to cover multiple tiles (0-1, 1-2 etc in UV space) and used a Composite texture node, along with the Offset controls of each referenced bitmap node, to assemble the individual source textures (bitmapDisp_u4_v2 etc) into a single "mega texture" (bitmapDisp).
Instances of this mega texture could then be plugged into a material in the same way as a traditional single-map texture (Fig.04).
Once this was done we set up a fairly straight forward mental ray scene with area lights and Final Gathering. We ignored passes on this occasion and just rendered out a final render of Qev with an alpha channel in order to composite him into a fairly abstract photo of an urban environment taken in the area around our studio, which is in Old Street, London, England (Fig.05).