State of the virtual art
At this year’s Develop, we asked three VR pioneers to tackle the issues facing visual design in this new medium
We gather a panel of pioneers to tackle the challenges of VR art
Gathered for the Edge panel entitled Next-Generation Visuals: Creating Art For VR, leading developers joined us at Develop in July to discuss the challenges facing artists working with virtual reality. In this edited transcript, SIE London Studio executive producer of VR Brynley Gibson, Rebellion head of digital Matt Jeffery, and Guerrilla Games Cambridge principal artist Shawn Spetch consider the potential and pitfalls. What challenges face developers creating art assets for VR games? Shawn Spetch We now have to think about things in terms of performance. So instead of just building a game like Killzone Shadow Fall, where we populate levels with a lot of detail, we have to scale that and we test performance while going into production. We optimise while we’re working, instead of putting a bunch of stuff in and then going backwards to make sure it fits into the framerate – framerate is the most important thing for us right now. We’ve had to rethink so many aspects – we use bold shapes, we make sure it’s readable, and we focus on believability. How do those aspects marry up? Brynley Gibson At London Studio, we talk about believability over realism always. So even though we use kind-of-realistic textures, it’s all about the shapes and we don’t go for lots of high detail. Battlezone’s a great example where you’ve got solid shapes, but it’s a style that doesn’t have a lot of noise in it – it helps people to not be overcome or bewildered as they move their heads around. Detail is not necessarily needed for great, believable VR experiences. Matt Jeffery VR should be a wonderful experience for everyone. We know that there’s a small percentage of the population who will always get motion sickness whatever we do. But for the vast majority, trying to make the experience as comfortable as possible is very important. SS We’re always trying to beat that 60fps because you have to in VR – there’s no way that you can drop frames. There’s not much of a division between design and art now – we actually work together as a unit to make sure that it’s all running within the performance target and we check GPU all the time. MJ It’s like Fight Club: the first rule is framerate, and the second rule is still framerate.
“When we’ve ramped up the resolution and textures, it doesn’t make great VR. It’s definitely not right”
How will this change in the future? BG We have to be cautious about this; we’ve found when we’ve done tests and ramped up the resolution and textures, it doesn’t make great VR. You think, “Oh, it will be brilliant, it will be crystal,” but it’s definitely not right. It’s about picking the right art style for your game, but also for VR. And there are many ways that you can do that – it doesn’t mean that it all has to be arcade-style. We’ve all been trying out different things, and that will continue for the next few years. And I think a language will develop. SS Some of the things we’ve tackled include reducing surface detail on textures. We watch our highs and lows, our darks and lights, as well. For example, blacks are not always the best thing to use because they’re harder to read – players can perceive them as a hole in the environment, or they might not be able to sense the specific depth between them and that area.
MJ When we make our more traditional games, like Sniper Elite, a lot of it is pushing the GPU as hard as we can. To use a bad analogy, you dial everything up to 11. When we did our first VR prototypes we did the same, and then when we put on our headsets it was overwhelming. It’s too much, and your brain doesn’t like it.
Will there always be a compromise?
SS We’re trying to push a photorealistic look for RIGS. For example, the sand in our Dubai map has a lot of visual detail in terms of [the grains’] waviness, but we use an inverse detail map in the shader, almost like a set range, where we’re drawing detail farther away from the player, and less when they’re close.
Polygons are probably our biggest budget consideration right now, so we’re trying to figure out ways to drop counts – reverse LODs are something we’re looking at. We take the fourth – or last – LOD and work backwards, which is different from just making a model in ZBrush and then decimating it. So we put that LOD in the map, and then look at the form and silhouette.
The reason we do this is that with the head-mounted display, the LODs tend to drop off with the field of view really quickly. That’s going to be representative of the distance that you’re going to be drawing most of the time. From there we can upscale and do the rest of the LODs down. As time goes on, you’re going to see techniques like that being done across the board, and people trying different things, and it’s just going to make things smoother.