Creating Stylized Art in a PBR World
With the use of PBR (physically based rendering) and PBL (physically based lighting), video game engines are able to render amazingly realistic worlds and characters. I have worked on a wide variety of titles since I started my career back in 2001. A sizable portion of that time went towards creating hand-painted stylized art. So, when I began to first make the transition to using PBR materials, I struggled to get the stylized look I was after. More often than not, models would not read as well as I had intended. The shape language wasn’t coming through, most of my texture details were getting lost in the more accurate reflections, and as a result, my work read as neither stylized nor realistic. It just looked broken. When I admitted that I had a problem, it was time to determine the causes and look for solutions.
Shapes and Silhouette
Without having the details painted into the texture, I noticed that I wasn’t getting as strong a read using similar proportions in my stylized models. After a number of tests I started to see results that were closer to the look I was after. Exaggerating the proportions more gave me a stronger silhouette with more of a stylized feel, and allowed for more information to be displayed in the normal map. When the models were less chunky, I needed to remove details or make them smaller to fit in the UV space, and this detracted from the look.
Details
This one was a little more obvious than the others, and in hindsight it feels like I should have noticed it from the start. Applying macro vs. micro details has been an ongoing learning process for me. As an artist, it is difficult to put the pen down and stop noodling on the details. Be they hand-painted textures, high-poly sculpted height maps, or fully-procedural materials, too many small details detract from the stylized look. They create a noise frequency that is too high, and they distract from the key shapes of the model. When a small amount of finer detail is used in combination with larger areas of rest, the overall model is much more pleasing.
A Question of Roughness
This was the area that I took for granted. I had assumed that it would work as long as the values were close to the material it was made of. I quickly found out how values that are too accurate tend to look too realistic. Go figure. For a lot of my materials, the roughness value is much higher than its real-world counterpart. Less reflection allows for more color and a hint of brush strokes to come through in engine while still receiving lighting well. This was still enough range to also be able to hint at surface details and sell the surface type to the player. Just like with height maps, too many small details took away from the look. Adding some subtle medium- and small details in that weren’t part of the height map was a good way to balance the lower detail level on larger surfaces that felt too flat.
To Metal or Not to Metal
This one comes down to personal choice and the art direction of the project. For me, since I tend to lean on the higher roughness side for materials, having true metals is a very strong contrast to the dielectrics in the scene. Excluding chrome surfaces, a dielectric with a low roughness usually reads very well as a stylized metal. This is even more true if you are trying to achieve a fantasy-style colored metal.
Lighting at Last
Lighting is a very involved topic and worthy of its own article. Final lighting often comes in near the end of a project. However, for creating truly effective materials, it is incredibly important to have a solid base lighting or, even better, a separate test map with neutral lighting. This will allow you to accurately view your materials in a more lighting scenarios. Since stylized art tends to be more saturated and use different values than real-world materials, we can’t look at the physically accurate numbers posted online. They can be a great starting point, but more often than not you will want to adjust them considerably to dial in your look.
In Closing
One of the most exciting and challenging parts of creating art for video games is finding ways to work within the limitations of the technology to create visually-interesting content that does not hinder performance at runtime. These limitations are constantly lessening as technology improves, but there will always be challenges, experimentation, exploration, and risks to be taken. This is what has kept my career exciting and fulfilling for 18 years. I hope that some of this will provide perspective on your own work: art is a great journey, and the more we share, the more we all can grow.
