Where is the Character Art pipeline headed? This is a question that I have been stewing on over the course of the development of my latest original character, The Serpent Templar.
Over the course of my career, the character development pipeline has had two major evolutions. The first big shift came to light when normal maps became an industry standard. Suddenly character development was driven by sculpture or scans, and in this way, the once all-encompassing game model was reduced to arbitrary topology. As rendering engines in games went on to embrace physically based rendering, we saw another evolution where realistic texturing became somewhat standardized across engines and the need for a wider array of baked bitmaps would change how we author our textures. Today, GPUs approach the ability to fully utilize a PBR rendering environment with ray-tracing in real-time and I am wondering, what will the next evolution in character development look like.
In pondering the future, I find myself fixated on a remarkable new Unreal Engine feature, nanite geometry. Somehow, the rendering engineers behind the Unreal Engine have empowered world artists to import static meshes in the millions of polygons. Something that literally seemed hysterical only a few years ago. Needless to say, this has had a profound effect on the way static assets are to be authored going forward, where PBR textures must be applied to the high-resolution model.
Suddenly, foundational surfacing features like Normal maps and Ambient Occlusion maps become obsolete in the high-end game rendering. This is because real geometry can substitute Normal/displacement information and real-time Ambient Occlusion has all the physical geometry it needs to render true. Remember, in PBR environments, the primary duty of Ambient Occlusion maps is to provide AO to forms presented by the normal map where there is no geometry to drive our real-time AO. So, at peak PBR rendering, a nanite mesh with no normal map won’t really need the AO maps either. So, you might ask, now we simply use two fewer maps in our materials, so what? Well, from a content authoring perspective, much of the detail generated in Substance Painter is derived from the hypothetically obsolete Normal and Ambient Occlusion passes. Taken further, Curvature masks are often generated from the normal map to create edge wear. World space normal maps can be used to pinpoint the underside of objects where dust gathers. Substance Painter is the industry standard in surface authoring for games and its foundation is built around the baking of high poly models. Smart materials rely heavily on your standard Normal, AO, cavity and, direction maps but now we need to author directly on the high poly models themselves. Specialized as it is, any artist can attest to the fact that material authoring on even highly optimized models within Substance Painter is less than smooth. Complex surfacing on a model in the millions of triangles starts to present a real performance issue. So you can see how the growing need to author directly on the high-resolution sculpted asset is starting to upset the apple cart a little.
Video Cards will get better and faster with time, but nanite is here now. Currently, to keep things running smoothly, artists are opting to work on a proxy of the nanite sculpt, aka baking the details into a lower subdivision to keep things snappy and provide something to bake those handy normal and curvature maps which drive detail. I just wonder, is baking this proxy geo really the ideal workflow? Or is it a workaround, given these are just the tools we have today? There was a time before Substance Painter when we authored PBR materials in Photoshop and this was our workflow at the time. I think we can agree that today, any PBR map editing done in Photoshop would certainly fall into the category of a workaround in lieu of a proper tool.
So as the need to directly author high-resolution assets increases and the need for bake maps decreases, what becomes of our art pipeline? What tools do we truly need to provide models of this fidelity? Well, we need an authoring software that can handily perform under the stress of tens or hundreds of millions of polygons. Well, we have that. It’s called Zbrush. But it needs to be able to isolate curvature, smoothness, and other surface detail masks from the high res model. Ok no problem, Zbrush has this as well. We also want to support procedurally generated patterns and weaves to drive surface detail. Zbrush checks that box too. So what are we missing then? Well, a native PBR environment would be nice and some more sophisticated layering would be appreciated too. It seems like a lot but I can tell you, after working on this project, the Serpent Templar, where I have textured the entire figure in its full resolution of 130 million points, Zbrush isn’t ideal but also isn’t too far off.
We are entering a new era with Zbrush with Maxon at the wheel. I don’t know in what direction they are going, but I feel like I see where asset authoring is headed and I firmly believe that Pixologic’s game-changing software has an avenue to shake the foundations once again. Nanite hasn’t permeated into character art just yet. As of this writing, deformable meshes and nanite do not mix but that hardly suggests that characters will miss out on this awesome tech. As an artist who would love to author every detail on my high poly sculpt, I certainly hope nanite enters the character pipeline sooner than later. I feel like I can see the path forward. Sure, deforming a nanite mesh seems like a tall order indeed, but what about character elements that don’t deform, like plate armor, or rigidly bound hard surface models like Guns or Vehicles? Somehow this doesn’t seem far off to me. Not in a world where nanite exists in the first place.
Can Maxon take Zbrush into the material authoring game? Can Adobe and the Substance Suite adopt a voxel-based approach that lends itself to nanite authoring? Perhaps, the very capable Marmoset Toolbag can step in and take on this challenge. They certainly do a great job of replicating most modern engine features. With nanite technology very much in its infancy, we still have some time to figure it out. As a game artist whose first love was traditional sculpture, I can honestly say that I am delighted to see where things are going.
Follow through to Art Station for more renders and other works