December 16, 2019
Survios shares valuable VR tips through the development of Westworld Awakening
In the game, you’re a defenseless victim who has to outrun a deranged android killer. In our interview, Survios Lead Producer Michael Patrick Clark, Lead Technical Artist Kevin Andersen, and Project Lead and Lead Engineer Daniel Zeligman talk about how they were able to come up with Westworld Awakening’s fear-inducing gameplay loop. With the game being arguably the best looking VR title on the market, the trio also talk about how they created the game’s stellar lighting, shading, and particle effects. Westworld Awakening also pushes the boundaries for character models and digital human faces in a VR game, and the developers expound on how they were able to get them to look and perform with the utmost realism.
Despite the high fidelity graphics, however, the game runs well. The developers also explain how they optimized their experience for VR’s more limited constraints and talk about how they made a truly interactive world that’s tailor made for the burgeoning medium. Survios was secretive with the development of Westworld Awakening, and its launch surprised a lot of people, especially considering how well the game has been received by players. Can you talk about how Survios’ relationship with HBO and the Westworld franchise began and discuss how involved they were with the project?
Lead Producer Michael Patrick Clark: HBO Interactive has had a long interest in virtual reality, and had previously tested the waters with Westworld and VR in Westworld: A Delos Destination. They wanted to advance storytelling in VR and sought out Survios based upon our pedigree of VR innovation.
HBO brought a tremendous love for the art and craft of storytelling to the partnership, and our two teams built a writer’s room together on-site at Survios. After a short time working together, both groups realized that we were really onto something special, and together we expanded our initial experiments, making Westworld Awakening everything that it is. That, in combination with the creative involvement of Westworld creators and producers at Kilter Films, helped us craft not just a short experience but a full narrative that sheds new light on the mysteries of Westworld and the Delos Corporation.
Being an android host within the Westworld universe who begins to break free and gain agency only to repeatedly die in an almost Groundhog-Day type scenario seems like the perfect narrative fit for a VR game. How was the premise for the game established?
Clark: From the beginning, we knew that we had to focus on telling a story that could only be told in the universe of Westworld. We started with some immediate goals. We knew we were going to take place during the second season, and that we wanted to expand the world by viewing it through a new set of characters. Our Narrative Director Bennett Jobling and Senior Writer Hillary Benefiel pitched several ideas, and we challenged and refined them until we finally concluded that this was the strongest, most interesting story that could be told. When we started seeing how powerful player reactions were to the beginning of the story, we knew we had something special.
Considering you play as a helpless victim throughout the game with no means of stopping a deranged stalker that's out to get you, the game expertly blends horror, adventure, and story-telling moments together while elegantly weaving it into the larger context of the show. How did Survios come up with and execute on Westworld Awakening’s gameplay loop?
Clark: Survios is an independent studio making multiple games; in fact, our pirate game Battlewake launched just three weeks after Westworld Awakening. We operate with fairly small core teams, augmented by multi-project team members that move from project to project as needed to quickly scale during production once an element has been proven through iteration and testing.
With these smaller teams and a collaborative environment, it’s vital that we have an excellent team, as we put tremendous faith in them to operate without a lot of handholding. With this trusted team and open communication, we’re able to quickly iterate and hone in on the best experiences by removing a lot of the bureaucracy that can come with large teams.
Ultimately the most important part of maintaining that cohesive, expert blend is reviews and testing, [and] quickly bringing the collaborators together and proving things out.
Westworld Awakening features top-tier lighting, shading, and particle effects, which make it arguably the best-looking VR game on the market. How did you incorporate these visual effects?
Lead Technical Artist Kevin Andersen: We edited the lightmass settings with DefaultLightmass.ini to ramp up the fidelity of our baked lighting, and made liberal use of emissives. Volumetric fog is used basically everywhere, relying almost entirely on the static volumetric lightmaps for shape and color. The custom depth pass was used to mask out the specular response on the floor to ground darker objects with a fake reflection underneath them. Small objects, for which a high-resolution lightmap would be wasted, were set to use the volumetric lightmap instead, while still casting a static shadow on their environment; a compromise that spared us a lot of baking time and lighting seams.
The game features an extremely detailed world with numerous interactable VR objects. How did you approach designing the environments in Westworld Awakening?
Clark: Environments are quickly blocked out by designers, tested for pacing and gameplay, and then our environment team takes a room to final quality art. Certain common props are turned into interactive pickup objects like bottles and the obligatory cowboy hat to introduce the idea that the player can and should interact with the environment. This isn’t a terribly unique process, but it is important.
But after that, we do something fairly unique: our narrative designer comes in and does additional set dressing, giving every room a unique narrative with additional propping. When these rooms have a narrative, then we write and place interactive narrative elements like newspapers, memos, e-mails, and notes. And because we’ve trained the player to interact with so many objects, it feels very natural and non-invasive to pick up and interact with a narrative object.
The game features best-in-class character models and faces. How did you get them to look so lifelike in VR?
Andersen: We use the forward renderer for its superior baseline performance and ability to use MSAA. It lacks a few of the features that make the deferred renderer UE’s flagship, so we made a few bits of tech ourselves, to achieve the level of realism that HBO wanted. For realistic skin in forward-lighting, we developed a lightwrap SSS shader, which we first used on Creed: Rise to Glory, and then further refined for use on head-scans in Westworld Awakening.
Lacking contact shadows, the eyes would have required too much shadow-map resolution to look right, so shading on the eyes was an animated lobe of fake shadow that used Blueprints logic to keep itself oriented to the head and eyelids. Unreal’s awesome Material curve feature allowed that fake shadow to also accurately follow a character’s blinking and squinting by simply reading the animation curve from within the eyeball Material. That same feature allowed the ambient occlusion inside the mouth to darken and lighten with the character’s speaking animation. Since we are dependent on static lighting that does not self-shadow on dynamic objects (like characters), the mouth being an animated cavity caused it to always be either too bright or too dark, which hurts realism when the character is talking. The rigging artist added a function to the character rig that just tracked the distance between the upper and lower lip as the character is speaking and output an animation curve with that information. Playing that animation back allowed the Material to lighten and darken the mouth realistically to avoid distracting light-leaking, as seen below.
Clark: It starts with crafting the right performances. As this was our first foray into high fidelity performance capture, we relied heavily on incredible partners. Our Voice Over and Motion Capture Director Kris Zimmerman Salter was a tremendous asset, helping us cast and capture some amazing performances from a slate of fantastic actors and actresses. We worked with partners at Snappers Systems and 3Lateral to develop our lifelike facial rigs, and our partner Cubic Motion worked with us to transform our performance capture data to in-game scenes.
For both pre-production stub-outs as well as for late pickups, we have our own Xsens mocap suit and Facewear Technologies HMC system that we used in-studio to ensure quick data capture and turnarounds with our internal animation team.
After we have some fantastic performance motion and voice, it’s up to our character and technical teams to ensure that we have appropriate Materials, blend shapes, animated normal maps, etc. that ensure our characters are capable of fully expressing the emotion that our actors and animators have worked into every element within Sequencer.
Westworld Awakening not only looks great but runs well, too. How did you optimize the game to run so well in VR?
Project Lead and Lead Engineer Daniel Zeligman: We had all of our talented artists and designers just drop all their assets in each level and said… ship it!
In reality, after shipping multiple VR games across all platforms, we knew going in we needed to establish firm budgets and benchmarks for all disciplines as early as possible. For example, we established our character vertex and Material count limits, our texture resolution and overall memory budgets, and our general limits on game simulation during pre-production.
However, having budgets is not enough to keep things running smooth. You have to continually monitor and adhere to them. For every significant milestone, the engineering and technical art team took major passes evaluating where we were “bound” in a given level’s section. We made heavy use of the Unreal’s profiling tools to identify threading stalls during asset streaming, and the general state of the game, render, and GPU threads. As an atmospheric, narrative-driven game, we were typically always bound either by overdraw, draw calls, or vertex count.
Our technical artists did amazing work developing our glass Materials to maintain the look and feel of the Delos lab, while limiting the impact to overdraw on a forward renderer. Our environment artists used instancing, merging, and atlasing across the entire game to reduce our draw calls and vert count. Lastly, we relied on precomputed visibility volumes, culling volumes, and custom gameplay driven actor visibility volumes to get us over the finish line.
Andersen: On the GPU side, we made sure to strictly limit the number of overlapping dynamic lights and only use high-quality reflections where they made the biggest impact. Since dynamic shadows are required for high realism, it was important to make them as cheap as possible. We did this by making low-poly shadow proxies that conform to the rendered object and only exist to cast shadows. So characters with 30k triangle bodies and eight Materials can cast shadows that only cost 2k triangles and a single draw call to render. Absolutely anything that did not outright require a specular response was made fully rough (removed specular completely) to save pixel cost. The resolution settings on the volumetric fog were brought way down from the Unreal defaults, because a localized haze can be achieved with a fraction of the resolution required for sharp volumetric shadows and light beams.
How big was the development team and how long did the game take to make?
Zeligman: The core development team was around 12 people and scaled up to around 35 in peak production over a period of 18 months from pitch to ship.
Survios has developed several VR games with Unreal Engine. What draws the company to using UE4?
Zeligman: The Unreal Engine’s expansive feature set along with source-code access allows Survios to get up and running quite quickly. The plugin-based architecture, strong baseline animation, visual effects, Blueprints, and core systems greatly speed up iteration. Epic’s continual support of existing and new features allows Survios to focus on making great games, and not engines.
Did the team have any favorite UE4 tools or features for Westworld Awakening?
Zeligman: Westworld Awakening, as a narrative-driven game, relied heavily on the use of Sequencer. We used it for all our “host control” gameplay segments and other scripted sequences. In addition, we relied heavily on volumetric lightmaps and lighting scenarios to set the stage of our atmosphere. Hank Harper, our terrifying villain, was developed through significant iteration relying on heavy use of Behavior Trees, Environment Queries, and the Visual Logger.
Can you talk about how Survios has leveraged source-code access?
Zeligman: Survios has been using source-code access since the pre 4.0 version of Unreal Engine. For Westworld Awakening, we made heavy modifications to minimize hitches during streaming, expanded the use of lighting scenarios to allow swaps at runtime, [implemented] Sequencer modifications, and a variety of other changes.
At this point in time, there are over 1,200 changes to the Survios’ version of the base engine. It would not have been possible to make a title like this without source-code access.
Survios has released high-quality game after high-quality game. How does the studio manage that as a company that solely focuses on VR, a new medium in which many companies are still trying to find their footing?
Clark: We’ve constantly tried to challenge the limits of VR, and each of our games represents technical innovation and new tools and technologies for us to incorporate into our next game. This is intentional, we have a goal of focused innovation. By building a game around a particular goal, we’re able to deeply develop that new system, and then incorporate it into all our future titles. Sprint Vector, our parkour game, proved out our Fluid Locomotion technology that drives Westworld Awakening and the upcoming The Walking Dead Onslaught (TWDO) will also feature brutal melee combat built from the Phantom Melee system we built for Creed: Rise to Glory.
Ultimately, though, our greatest strength is the team itself. We’ve been able to build and retain a talented, fantastic team, and we invest in them. We have a mix of experienced senior developers and bright young people, many of whom have come up through our strong internship programs. We are dedicated to maintaining a good work/life balance, which is especially critical when you’ve got shared teams like our character and audio teams that move from project to project in rapid succession.
This is the second game from Survios that is based on an existing movie or TV show tie-in, alongside Creed: Rise to Glory. Are there any tips you can provide to making a good game based on a popular existing IP?
Zeligman: Working with any existing intellectual property can be a challenge. The most important items I recommend for any team working with an IP are the following:
- Establish your approval process with all stakeholders as soon as possible.
- Ensure all team members understand the licensing constraints.
- Don’t be too literal. You are making a game that is in the spirit of an IP, not the IP itself. Don’t let real world or IP constraints keep the player from having a satisfying and accessible experience.
Thanks for your time. Where can people learn more about Westworld Awakening?
Zeligman: Westworld Awakening is out on Steam, Oculus, Viveport and various VR arcades around the world.