Overwatch HQ: Creating A Game Level for Portfolio

Worth Dayley discussed his enormous project, where he builds an amazing Overwatch HQ location in Unreal Engine.

UPD: If you’re interested in using the Overwatch brush, which Worth created, you can download it for free here. 

Introduction

My name is Worth Dayley. Growing up in Orange County, California, I brought Worldcraft home from my friend’s house on zip disks and started making maps for Quake. I’m 7 years into my career in the games industry as an environment artist initially on every iteration of Disney Infinity (Toybox, Inside Out, Finding Dory) and later as an art director at a couple startups for mobile and AR/VR development. I’ve also been lucky enough to work on the past 2 seasons of Rick and Morty as a 3D artist. Currently, I’m laying low in Salt Lake City, on a portfolio hiatus.

Overwatch Project  

Looking at senior-level environment artist portfolios, you see a lot of simple screenshots of completely realized environments. Looking at my portfolio, I saw a lot of assets and texture spheres. Two of the companies I’ve worked for in my career didn’t exist long enough to ship a product, so I’ve felt the frustration of collecting work experience without anything visually to show for it; now I’m diving deep on a long-term project and looking to join a more stable, gameplay-focused developer. I decided to produce a fully-realized, professional, shipped-quality environment based on the Overwatch art style to bring my art up to where I wanted it to be.

I talked with some friends in the industry, and they were big proponents of finding an expert piece of concept art and building from that, but I’ve always tried to inject my own visual design into my projects as much as possible. Besides, where was I going to find unused Overwatch concept art? This was right when the official Uprising comic came out, showing the inside of the Overwatch headquarters for the first time and my friend suggested I do something along that theme, which turned out to be a great brainwave. The comics were essentially official, approved, unused concept art, but there was still a lot of gaps that I would have to fill in with my own visual design. 

From the outset, I wanted to be extremely mindful of what I call “creative friction”, which is the opposite of momentum. Creative friction comes from anything that makes it easy for me to not work; disorganization, life stuff, even having Facebook open on another monitor. The more things that have to be fixed or put into place before I start working, the less likely I am to start working, and the more the friction builds up and pushes against progress until it kills the project. One thing I’ve focused on a ton in my OWHQ project is reducing creative friction as much as possible through good project management practices like writing down and following a proper workflow pipeline and setting up my project’s organization and folders ahead of time in logical, clean ways. So far it’s been paying dividends: I’ve minimized the time I spend trying to find files or fix errors, and when things do distract me and kill a day, it’s much smoother to jump back in to my work. 

Additionally, I feel that being aware of what goes into creating aspects of the project in other departments only makes me a better production artist. I call this having a full pipeline mindset in your work. When I first started as an intern, I would get a task, I would perform that task, and then I would hand the result off to my mentor or art director and then get a new task. This is fine for an intern, but as I progressed, I started to collect as much information as I could about what other departments were doing. I eventually learned what happened to my assets after I handed them off and actually learned how to take my assets several steps further than I was expected. I learned how to trace the path of my art from my workstation all the way to integration within the game. Most importantly, I also learned where my decisions impacted the lives and workloads of other people down the pipeline. Eventually, I felt that I had just clear enough a picture of what the broader consequences of my creative decisions would be. Do I put text in this texture? That will have to pass through legal and localization. It’s likely to get changed or cut. How can I design around this? If I model and texture a digital readout into this machine does that have to go through a designer to give it functionality? Rigging to animate it? FX to make it prettier? Sound to support it? How many bugs will all of these moving parts create? Is the bang for the buck worth it? Who do I have to talk to or inform? I stepped on several toes in my career before I wised up to this butterfly effect, but these were lessons well learned. Since then I try to be mindful of and to dabble in multiple disciplines.

Blockout 

I view big projects like this as a series of creative problems that require solutions. I tried to start with the biggest, broadest problems and work my way down to the smaller, more detailed issues.

After finding the concept art and deciding what my playable spaces would be (interior hallways and rooms, exterior terrace, etc) I focused on the overall picture. Where, exactly, was this headquarters facility located geographically (what country/region)? Where was it located specifically (in a city or forest)? And what did the overall facility look like and how did my playable areas fit into that? The scary and laborious part of environment art is that you get absolutely nothing for free; you have to design and build the pebbles on the ground and the sun in the sky and everything in between, so I really needed to have a clear vision of what and where this place was. 

I like to brainstorm these early concepts away from my computer. I went for a walk for a couple hours and talked myself through it, deciding to set it in the Swiss alps with access via a futuristic gondola and a sprawling alpine valley below with the Matterhorn visible in the distance as this vision seemed to fit Overwatch’s comic book hero motifs. I scribbled terrible messy sketches in a little sketchbook I had brought with me right there on the sidewalk as I pieced together hangars, hallways, office blocks, and gondola landings, and by the time I went home I had a reasonable layout in my head. I really tried not to linger on these early steps much at all but move on quickly when I had enough figured out.

Next, I went immediately into 3D. Sometimes I’ll do concept drawings instead, but my goal here was to get into the game engine as soon as possible and start running around the space. “Sense of place” and how it felt to move around the world in-engine were the most important things to me in this step. I started laying out my rooms with cubes and boxes and importing them into Unreal Engine 4 and running around them until it felt ok to move around the space.

I wouldn’t really call this step blocking, it’s more like underdrawing. Underdrawing is when an artist lays down light structural lines to serve as a guide for a drawing. It’s not good drawing, and it’s not meant to be seen by anyone except the artist, but it’s where they solve a lot of big problems before they even start putting down real lines, and this is what I use my initial modeling for. I solve some spatial problems by running around in it, and then I use screenshots of the simple blocking to do concept drawovers where I solve the actual design problems such as implementing the visual language of the style or where to put pipes, boxes, story objects, props, etc. This way as I draw my concept, I know that what I’m drawing is screen-accurate and will fit in that space. 

After this, I began experimenting with a pipeline, writing it down as I developed it. Then I chose a contained area, in this case Strike Commander Jack Morrison (aka Soldier 76)’s office, to use as a vertical slice to push through my pipeline to a level of completion. Doing a vertical slice helped me figure out how I wanted to treat project and scene organization, modeling, texture authoring, materials, lighting, level of detail, post processing, and other aspects. These initial steps: underdrawing modeling, paintovers, and vertical slice, are where I learned most of what I needed to know to speak the language of the project and it’s where I wrote my pipeline and set up my project organization. This ground work took a long time, but it was vital work.

Details

Doing any kind of style emulation is a real exercise in observation, particularly trying to see what is actually there versus what your brain just thinks is there. I spend a lot of time alone in my own custom Overwatch games staring at walls and props. Without knowing the minds of Blizzard’s artists, I think I can say that any given element or asset in Overwatch should (A) only have as much information in it as it needs to read as what it is and (B) not call too much attention to itself when it’s in context. One of the first things I did, since this was a portfolio piece, was to use higher resolution textures for various elements, and I noticed that it actually detracted from the Overwatch-style look! Sometimes just cutting the resolution of a texture map in half was enough to bring a prop into line with the style because the higher resolution was drawing too much attention and importance to the object when compared to the rest of the scene. 

Scale is something Overwatch really plays loose with, and I think it all ties in to a philosophy that these environments are treated as illustrations more than collections of CG assets or recreations of real places. If a desk needs to be thicker than Genji’s neck to create a better composition, then they’ll do that. Chances are most players (except the ones like me who are staring at desks for minutes at a time) won’t care if things are slightly wonky as long as the composition they make up is beautiful. In my vertical slice, I was working on wall elements and giving them what I thought were really chunky bevels. I would get them into the engine and then look at them in context, though, and realize that what looked chunky in Maya wireframe still looked pretty conservative in-game.

I would have to push my curves and bevels in Maya to ridiculous degrees to get them to feel right when lit in game. That meant that some objects had to be scaled up relative to the player in order to model them with the right shape language, just to get all the elements in there. Tables get pretty thick and chunky when you’re beveling their corners so much that their straight planes disappear!  I did end up having to remodel a lot of the architecture in my vertical slice after I finally admitted that I was being too conservative, but vertical slices are meant to teach you these lessons, so mission accomplished! Overwatch definitely subscribes to that old illustrator’s mantra of “If it looks right but it’s technically wrong, then it’s right”.

Design

As an environment artist for video games, I’m interested in making video games first and making cg art second. For gameplay-driven games like Overwatch, and especially competitive shooters, the design of the maps directly influence the feel of the art. I wanted people to look at beauty shots of my map and easily imagine themselves running around and playing Overwatch in there, so I felt like I absolutely had to respect the design and gameplay rules that Overwatch maps followed.  There are spawn rooms, paths that you are meant to move and fight through and arenas that you are meant to fight and die in. There are also secondary and tertiary paths, lots of vertical space for mobile heroes, and vignette areas that are full of story/world building props (and health packs, usually). I tried to represent all of these in my project.

For example, when I did the underdrawing model for the exterior terrace, I initially just made a big, open space, but it felt way too open for Overwatch. I went to a couple maps that had some of the largest single open spaces in the game, like the areas right outside the spawn points on Oasis – City Center and I actually timed how long it took an average speed character to run from one side to the other. It took about 10 full seconds. Additionally, I noticed that while it took 10 seconds to traverse the whole space, I was slightly course correcting as I ran to avoid obstacles every 3-4 seconds (sometimes less). Rarely in Overwatch are you walking in a straight unbroken line for longer than about 4 seconds. This was fantastic data for me and informed my art direction on the terrace. I made sure the overall space matched that traversal time and I cut the space up with little cutouts with alpine trees poking up through into the terrace space, breaking up the straight lines.

Another great example is one of my main areas: the domed Atrium. I initially planned that room to be nothing more than a 90 degree corner to match the Uprising comic concept art, but I realized that all my interior spaces would be way too enclosed and almost all of Overwatch’s medium and large spaces are vertical to allow characters like Pharah to fly, so I went with the geodesic dome and expanded that area.

Materials

The textures and materials have been a real challenge with emulating the Overwatch style. The texture treatments create the stylized, illustrative look that we all love so much, and if you don’t carefully hit the right notes, then the song goes off key. In my observation, Overwatch’s textures follow a pretty clear hierarchy of read from most important to least important: Color, hand-touched painted detail, normal map, metalness, then finally roughness/gloss detail. Color is huge in Overwatch; not just for aesthetics, but for clarity of read and working together to create compositional harmony. Overwatch maps aren’t so much built as they are painted like an illustration.

Early on when I really started examining the maps closely, I noticed that Overwatch’s environment art utilizes a lot of really basic, simple techniques but it implements them in the best, most beautifully meticulous ways. There’s not a ton of fancy shaders going on and the maps aren’t packed with procedural doodads. They’re just composed of solidly modeled meshes with beautifully painted textures. After creating a space and deciding on a general overall color scheme, I tried to stick to this simplicity when painting my textures: 

• I would identify what material I needed to make, say a tile floor, and then quickly model a high res version of it off to the side of my scene in Maya. I’d bake that as a normal map and sometimes a curvature map to a flat plane in xNormal (remember that program?) and then do the rest of the texture in Photoshop.
• Occasionally a material would require some more randomized or organic detail like cracks or divots so I’d spend a few minutes in Zbrush quickly carving those out to add to the normal map.
• The basic albedo tends to be very graphical, letting the overall color carry the most importance. Often times just adding the curvature map to the albedo goes a long way.
• It’s very important to get hand-touched variation and detail into the texture whenever you can to get that hand-crafted feel, even if it’s just a quick pass. I made my own version of a Photoshop brush that they use at Blizzard that I’ve called “WDBootlegOverwatch” that I use on all of my textures.

• Next, I would look at what kind of roughness the material needed. Probably about 25% of the time flat roughness amounts differentiating between materials (like tiles versus the grout in between) work just fine. When more detail is needed (like on walls where you’d want the variation to break up the flat planes) I’d pull in one of a few tiling roughness maps that I’ve made by running photo reference through several Reduce Noise filters and touching up by hand. I’d overlay those on top, add whatever hand-painted details I had in the albedo to accentuate them, and then balance the levels until it looked good in-game.
• Metallics in Overwatch still rely heavily on the albedo and the hand-touched look. Some of my metallics have a less than 1 metallness amount just to push the dark mirror gloss look of Unreal’s metals back and let the albedo show through. I also play a lot with my reflection spheres’ brightness levels to pop the highlights and get a more stylized look in-game. What’s most important, though, is the simple read: Is this thing metal or not?
• Finally, I make ample use of decals. Overwatch maps, and especially the Overwatch facility art sets, have a certain greeble language which comes through pretty well with decals. Sunken screws, rectangular cutout bites, vents, seams, and metal grommets all go on top.

Restrictions

I really want to make games, and art is how I contribute to that process. I try to be a game maker first and an artist second. I think the fact that our art is meant to accommodate a player is the main element that differentiates game art from other CG art. I’ve made and seen many game art pieces where the space is beautifully rendered but I can’t see where I, the player, would fit in it. I look at the environment and I can’t imagine playing in it, and I think that’s a missed opportunity. In my OWHQ project, I wanted to make it easy for people to imagine playing in my environment.

I’ve seen some artists get really frustrated when they perceive that constraints like level design are acting on their creativity, but I’d counter and say that creativity is born in how you respond to and incorporate the demands of constraints. In my work life, I love the process of getting a greybox level from a designer and then learning what it had to be. It’s always either “This absolutely has to be a police station. I know it looks nothing like a police station” or it’s “I have no idea what this is supposed to be, you figure it out but don’t dare change the shape of the playspace!”

Forcing your art to merge with and support gameplay is not only just how it works in the industry in my experience, but it forces you to attempt solutions that you never would have thought of otherwise. The domed Atrium in my project and the side-path breakroom were both the result of me considering the demands of gameplay. The entire design of the terrace with the tops of trees poking up through cutouts to the mountain below came from bowing the demands of design. I never would’ve considered a lot of these elements, but they are now some of the most striking, defining features of my project.

For personal projects, picking a reference game that you either emulate or simply use as art direction/creative direction reference can help solve some of the more obscure problems you face and produce an end result that is beyond what you would’ve just thought up without any constraints whatsoever.

Key points 

In the process, I’ve been able to deep-dive examine a favorite and wildly successful game and test my instincts and my process against the masters. Instincts are just guesses without being tested, after all. 

I’ve learned a lot of things, but a few that are extremely important to me:

• Creative friction is the force that kills projects and you should acknowledge it and work proactively to reduce the problems that cause it.
• Ask yourself if you want game art to be a career or if you want it to be a day job. If you want it to be a career, then you need to look beyond just the modeling and texturing and start to consider the full pipeline: organization, project management, documentation, developing and writing down procedures for yourself to follow, respecting game design, and keeping your mind on the final result. You gotta ship something or no one will care.
• After everything is planned and in place, it just takes work. You have to show up and do a little bit or a lot of work over and over. There are few tricks to it, you just have to sit down and work on it until stuff starts getting finished, and then you have to keep sitting down and working on it until you run out of problems to solve or the deadline hits.

Worth Dayley, Environment Artist

Interview conducted by Kirill Tokarev