Don Arceta talked about the experience of developing Dylan (Anthem) while working at Bioware: environment design and production, assets, composition, challenges, and more.
Some time ago, Don also talked with us about material production. Read it here
80lv: Don, it’s been a while since we’ve last spoken. What have you been up to? What projects have you worked on? Were there any interesting things you experienced along the way?
Hello again! It’s been a busy past couple of years for me. I still play around in Substance Designer in my spare time but a few things have changed since the last article. I’ve hopped across the pond from Canada to the UK, going from Bioware where I’ve worked on the original Mass Effect Trilogy and Anthem to Playground Games where I’ve recently worked on Forza Horizon 4. Going from fantastical science fiction projects to an authentic contemporary project was a huge change for me and an invaluable learning experience. Recreating real areas of the world that we experience every day has its own unique set of challenges and is just as challenging as crafting distant worlds that only live in imagination.
80lv: You were lucky to have the opportunity to work on Anthem, which was first called Dylan. Can you tell us a bit about this experience and your main tasks on this project?
After completing the Mass Effect 3: Citadel DLC, the core Mass Effect team, myself included, moved onto Bioware’s brand new IP, Dylan (that became Anthem in the future), which was to be an ambitious next-gen, realistic, living open world.
To give some context as to where the video game industry was at that time, the PS4 and Xbox One were still almost a year away from release, procedural texturing was in its infancy, photogrammetry wasn’t a ‘thing’, and Frostbite had no PBR.
On Dylan, my main tasks as Lead Environment Artist fell into two categories which were the ‘what’ and the ‘how’. Working closely with the Art Director, concept art team, and design team, we explored ideas and themes to figure out ‘what’ the visuals of the game were going to be. What kind of tech was in this world? What types of biomes were there? What are the next-gen visuals? What makes our game visually different from Mass Effect? These were the types of questions we would try to answer during those early days.
Figuring out ‘how’ we were going to create the visuals meant ramping up on new technologies, investigating different workflows, and testing out pipelines. There was a constant collaboration with the Technical Art Director, graphics programmers and tools programmers.
Frostbite was new to us. Speedtree was new to us. We were transitioning from 3ds Max to Maya. Should we use Substance or Quixel? How can we make the world dynamic? There was a lot for us to figure out and this meant a lot of trial and error.
Designing the Look of the Environment
80lv: What good approaches to test the look of the environment do you have at your disposal? How do you look for that necessary design?
Once we received a concept from the Art Director or concept artist, we would start blocking out the main elements, shapes, spacing, and composition. We called this the whitebox phase. Too much detail during this phase could be too distracting, so whitebox mesh was kept simple to keep the composition focused. Another reason we kept the mesh simple was to allow for big changes to be made quickly and easily if necessary. If we were struggling with the whitebox we’d sometimes get paintovers from the concept art team. Once we were done whiteboxing, we would start to look at a more detailed reference. A lot of the reference was gathered by the concept art team and Art Director, however, the environment art team would also gather their own reference as they needed. We had different kinds of reference, some were inspirational while some references were more authoritative. There was a lot of different references we explored during pre-production. NASA, Calatrava architecture, Moorish architecture, Southeast Asia, and Hawaii are a few examples of the variety of references we explored.
80lv: Do you rely a lot on the different existing libraries of objects to do a quick look of the space? Do you happen to utilize Megascans and Speedtree and if yes, how do they help?
During those early days of Dylan, to build up an area we authored assets as we needed them. There were opportunities to use existing libraries of assets from other projects going on at Bioware but we ultimately decided to build assets ourselves. This was mainly for two reasons. First, was to establish a unique visual identity for Dylan and second, was to learn, practice and refine our authoring workflows.
Speedtree was one of those new workflows for our team that we invested time into early on. Dylan was going to be quite a foliage-heavy game so the procedural, non-destructive nature of Speedtree was important. Speedtree enabled us to quickly make variants, small tweaks, and large adjustments easily.
Photogrammetry, which was a rather new industry technology, was something we eventually started investigating in pre-production. We tested capturing a lot of different types of assets, doing a lot of scans in-house and around the studio. As that workflow began to tighten up we did a few photogrammetry trips out to Hawaii to help build a library of assets to use for our early visual prototypes. Megascans was still in beta at the time and was something we looked at briefly but ultimately didn’t use in pre-production.
80lv: What composition tricks do you use to build the look of the scene? Would be awesome to get insight into the way you think of lines and plan the composition!
Most of the time on Dylan, we would have pretty awesome concepts to work from but there were also times when concepts weren’t available to us. This is when I’d usually fall back on a lot of little compositional tricks that I’ve picked up while working on the Mass Effect Series. I like to keep things simple. Vanishing points, repetition, and lighting are hooks I usually turn to when trying to create an interesting composition. Strong lines I find useful in leading the eye to a vanishing point and I like to use repetition to sell the sense of scale and distance. Repetition also allows for opportunities to introduce contrasting elements to generate interest. I’ll use lighting to accentuate the strong lines and help guide the eye and sometimes I’ll use lighting to create or bolster a focal point. Finally, when taking a screenshot, I like to frame the camera using the rule of thirds to tighten up the composition that much more.
80lv: Some of your landscapes look absolutely stunning. Could you talk a bit about the way you usually experiment with these landscapes and how you build them?
With Dylan being an open world game there were many tests done using Frostbite’s terrain systems. Our Technical Art Director was developing the idea of terrain brushes or stamps. The goal was to be as non-destructive as possible. We wanted artists to be able to place a mountain as a ‘stamp’ and have the ability to easily move, rotate, or scale that ‘stamp’ at any given time. For these tests, I set up the terrain materials to be procedurally placed based on the slope angle of the heightfield. It was set up so that a tri-planar rock material would show up on steep cliffs, a grass material would appear on gentle hills and a scree material in between. Grass meshes, flower meshes, rock meshes, and other mesh scatter would procedurally spawn on the appropriate terrain materials. This procedural setup allowed for fast visualization and the terrain stamps enabled me to quickly iterate compositions. Initially ‘stamps’ were created in World Machine, but later on, we started to use real-world satellite data which gave us more realistic results.
What Makes Evnionrmnet Art Good
80lv: What way do you evaluate the direction that the environment art for the game is taking? Are you interested only in the look or do you also take into consideration the pipeline and workflow? It must be quite a responsibility since it will influence and lead other artists.
Good environment art isn’t just about how it looks. It is also about how it is made and how it can scale. Artists should have workflows that make it easy to author art at the quality, consistently. In Dylan, the player was essentially able to reach any spot in the world so this presented a challenge. We didn’t have the scope to hand touch every inch of the world so we needed a viable workflow that would enable us to populate the world with handcrafted quality details on a very large scale. During pre-production, this forced a big mental shift for our team.
Game Art: Challenges
80lv: What were the hardest things artistically and technically when building the spaces for the game?
On Dylan, it was important that visuals not only looked good, but also worked well for gameplay, animation, and were performant. To get the art to work with all the requirements of the other disciplines was very challenging. There was a lot of trial and error, compromises, and creative thinking. Since Dylan was set in a fictional world that we owned, we had the power to change and justify things as we saw fit to make elements of the game work together.
Building a Unified World
80lv: The final question might be a bit general but still relevant: what is the way we can make different environments in the game feel like a unified world? What are the tricks that can help you build the whole space?
There are many different ways to make a world that has a variety of environments feel unified. The simplest way to achieve this is to use the same subset of assets across the world. The subset of assets should visually work at variable densities and should be flexible enough to be placed in very different situations. Architecture and foliage are good examples of this and for Dylan, a specific example would be the ruins. Ruins could exist almost anywhere in the world and could be contextualized in unique and interesting ways. Ruins overgrown in a jungle would look quite different from scorched ruins in a desert but both would be just as striking.