Building The Martian-inspired Scene in UE4

3d artist Jack Parsons (The Beginner’s Guide, The Stanley Parable) talks about the production of his sci-fi environment in Unreal Engine 4. He gives the complete breakdown of the scene, talks about the production of the assets, materials and lighting.

Introduction

My name is Jack Parsons, I’m a British 21-year old self-taught 3D Artist. I worked with Galactic Cafe on The Stanley Parable and with Everything Unlimited on The Beginner’s Guide.

I started off by making custom maps for Source Engine games like Garry’s Mod and Counter-Strike when I was 14, making my first 3D models in Blender, before learning 3DS Max during some work experience with an archviz studio. In 2012 I was brought on by Stanley Parable writer Davey Wreden to be the main artist/level designer on what became The Beginner’s Guide. I got to work on Stanley too, though the majority of my contribution to it was actually the Minecraft part, funnily enough.

Last April I moved to Malmö in Sweden after finishing university (Computer Science), moving up to Stockholm with a friend last September. Since then I’ve just been working on art for my portfolio– I aspire to one-day work at DICE, or a similar studio in the city. DICE was my main inspiration that drove me to get into game art.

Quick warning: The following scene was made in a coffee shop, do not attempt the same unless you want to answer a lot of questions from passers-by!

The Martian Outpost

I started working on the Martian Outpost when Quixel Suite 2 came out in late November last year, and considered it finished mid-January. I wanted to start learning Quixel right away, so I imported a bunch of my existing models and had a good time texturing them with NDO and DDO, figuring out how everything worked; not long after, I felt that the best way to practice would be to start fresh make a full scene. I’d spent the past 12 months being obsessed with astronauts and the International Space Station, making loads of interior and exterior scenes of it in various game engines – so after watching The Martian like three times over there was no way I was going to let that inspiration go to waste!

Planning & Modelling

From the start I knew I was going to be creating quite simple geometry, adding most of the detail-work later on in the normal and roughness maps. My main focus was to use this project as a learning experience to see how much goodness I could squeeze out of using NDO to paint normals. I was surprised by how much I could do with it! The panels and the big metal parts could totally have been made in geometry, but this time around I valued the flexibility of using the normal map more.

To speed up the modelling process I just used edge loops to smooth over hard edges, instead of working with a highpoly and baking normals. Of course, this meant I ended up with a higher polycount overall, but for this scene it made little difference as most details were in the materials, not the mesh.

The two main floor models came first – I made them together to experiment with the sci-fi style, before I decided to turn it into a scene. What are those big black-and-white things, you may ask? I have no idea. I think in the end I settled on them being battery banks. Anyway, I placed a few of them around in Maya to get a feel for things, then sat back and thought about interesting directions the scene could take. I imagined a very wide combination of room and corridor with much less light toward the walls, and a central brightly lit walkway with either supports or hanging walls separating the two, holding the light in the centre. So I plopped the assets into Unreal, added some basic contrasty lighting that looked like an interesting start, took a screenshot, and made a quick & nasty rough sketch of my idea on top of it in Photoshop. The sketch was mostly a bunch of line-art – I’d show it here but I’ve lost it! The main additions were the big metal support struts, though they were a lot bulkier originally.

Next I used the sketch to separate the scene into individual assets, and figured out which ones were more important (so I could focus on them first). I then made the decision to model all of them in their most basic, untextured form, which took a day or two. I used this blockout phase to discover different shapes I could use – I wanted to give the scene a lot of depth, and especially wanted to avoid anything too flat on the walls and ceiling. To my surprise, many of the final meshes ended up being simple evolutions of the existing blockout pieces, which I think says a lot about what NDO can add. My modelling process generally began with figuring out a model’s dimensions, extruding out a bunch of interesting shapes, fixing smooth shading, then thinking about what the model’s function would be at the last minute. I still think I made the walls too flat though.

Posting my WIPs in the Unreal Engine 4 Developers Community on Facebook gave me a heck of a lot of useful feedback at this stage, highlighting parts I needed to change, either because their shading looked weird or because they didn’t fit the scene at all.

Managing the scale of the project occurred in the blockout phase, aided by a bit of sneaky visual cheating. Pretty much everything in the scene has some element of symmetry – not exactly ground-breaking but it does provide a bit more variation for little to no effort. I will say, the most important part of the creation process here was having some sort of plan. I didn’t start off with a plan, but once I had the general idea I made the sketch, planned which assets I would need, then made them one-by-one. Other less mission-critical props came in later and I was able to bring some in from previous projects, redoing textures where appropriate to match the wear and tear of the Mars environment. The planning meant I was able to finish this scene a lot faster and with a lot more ease than ones I’ve made in the past. Before, I’d plan too little and be crushed by a project’s scale, or too much, meaning I’d get too caught up in the details – and I’ve found that it’s even possible to do both those things at the same time.

A common problem with modularity and having a lot of symmetry is that the scene can become very same-looking all around. Every time I see something written about this problem, there’s usually the same emphasis: Lighting. And I agree. You can almost make the same room twice and people won’t notice if you change up the lighting enough (nicely discussed in this presentation on Skyrim by Joel Burgess). I think because this scene is just one large room, it’s quite acceptable for there to be a lot of repetition – if it were a whole complex that used the same assets all over, it would begin to be a problem, people would get lost in it and such. Keep in mind lots of real-world installations and environments have lots of the same thing in them, too. So when it came to making the individual details on each asset, something I noticed was I could make them look like they had a real-world standardisation to them. Pieces that look like they could be easily replaced helped sell the repetition as being more ‘genuine’, at least to me. Though the bigger stuff in the scene didn’t need to look entirely functional, I figured it was far more vital that they just looked like they could support their own weight. And in the end, a light dusting of props on top of it all breaks up the otherwise uniform silhouette.

The hardest part to design was definitely the airlock, the centrepiece. Once I had the design down, the modelling process wasn’t uniquely challenging in any big way, but I stumbled through several different designs before resorting to taking it directly out of The Martian. I fully admit to that.In times of great need, reference is your friend!

Materials Breakdown

Quixel was used to make and export pretty much all the materials you see. The way you start is by creating an ‘ID map’ from flat colours, to specify which parts of the model have different materials. My Maya script of choice for doing this is Nightshade Blockout. I then import the mesh and its maps into Quixel DDO and use their fantastic material library to apply materials to the various parts of the ID map on the model (sometimes after using NDO to paint normal map details and bake AO). You can adjust how much each material affects the bump and roughness, and add in things like scratches, dirt and rust on top. DDO does pretty much all the grind work for you, allowing you to experiment a lot more and spend more time telling the ‘story’ of each model – how it was made, what it’s for, and what’s happened to it over the course of its use. The idea of setting the scene on Mars came from me adding a light sand layer to the floor pieces and wondering what it would look like if I made it red!

With some models I wouldn’t bake out an ID map at all, and instead work in NDO first to draw in details and additional shapes (e.g. the large metal pieces with circular holes in them), then use that PSD to create the ID map, with the UV outlines overlaid on top to guide me. The texture resolution for each model was 2048 square – I would have gone up to 4096 but didn’t have enough RAM in my laptop for Quixel to use at the time. (I’ve now upgraded to 16GB).

I drew on a lot of past experience modelling space stuff when working with the materials, especially with the metals. Treated aluminium is used extensively on board the real-life space station for its low mass, with stainless steel taking its place in more structurally crucial components. As far as plastics go, again they’re very light and similar to the ones used on aeroplanes. A lot of my material choices were simply to create contrast, too – and that doesn’t necessarily mean colour contrast, it can mean choosing to use both metals and plastics, and mixing hard surfaces with soft surfaces.Fabric coverings here and there are to stop astronauts bumping their heads, but can also add a bit more relatability to the scene.

When I work on something based on a place or subject that many people have become familiar with, I think it’s essential that I understand it closely, why it’s built the way it is, how the place is used on a day-to-day basis and what the most important (or even the most respected) parts of it are.

Over in Unreal, my material setup was relatively simple. I had parameters for multiplying the roughness, AO, and emissive maps, as well as a colour input with a mask so I could change the accent colour (you may have noticed both blue and orange versions of the same model). One other multiplier I have is for the metallic texture – I’m not quite happy with how metallic Quixel makes my models look so I bump it up a bit. My metallic multiplier generally stays fixed at 1.2, with roughness hovering around 2.5.

Lighting

I’m a fan of having just a few, bold lights, instead of making the lighting more uniform. I tend to unconsciously follow the same three-stage pattern that I learned from some interior design book I read in a bookshop once. The first, most important lights highlight the functional areas of the environment – so in my scene, the airlocks are brightest. Second, I think about what lights would make the rest of the room appealing, finding positions that cast the most interesting or dramatic shadows, or simply illuminating areas that I find are too dark – and thirdly, when all the additional props and set pieces are in, I think about illuminating some of those directly too. Of course in a game environment, what would be functional to a character in that world might not be functional to a player, so a lit table would arguably exist in the third category, not the first.

To mix up the contrast more, I varied light temperature. Mars is a cold place, and I wanted to capture that feeling through light and atmosphere (which is why the mist effects are there, as Martian soil contains a lot of moisture). I always used the ‘Use Temperature’ setting here: I looked up a paper on ISS external lighting and found that they use a temperature of around 4200°K. My warm lights are at 4320 degrees and the cooler ones are at 11,800. They all use the default ‘Complex_IES’ IES texture from the engine files. Additionally, I threw in some coloured emissive parts to many of the environment models in places I thought would look cool, sometimes supplemented with spotlights to show up more in reflections.

When it came to the material aspect of lighting, I played around with roughness a lot in DDO to achieve what I wanted, using dirt and sand layers to subtly vary how each material reflects. I maintain a belief that how my materials interact with light can be as important as the lights themselves – so, when it came to positioning and angling each light, I wanted to make sure these interactions could be shown. Something I’ve found is if I put some surfaces in the dark, but still in view of a light source, I can create quite dramatic reflections and interactions. It’s obviously not an exact science – just a matter of experimentation with light placement. Using Unreal lets me do that quickly and flexibly, thankfully.

Meet the Props

The Outpost needed props to give it life, but not all props are created equal.I consider a prop to be more important when it’s used more frequently in the level, or used functionally to tell a story. In preparation for prop placement, I bolted steel rails on most parts of the habitat to accommodate the blue handrails and a few other things, simply because I see them everywhere in NASA imagery. On orbit these rails are used to attach everything imaginable, and no doubt will be useful on Mars in the lower-gravity environment. I already had the blue handrails from a previous project of mine, but gave them the Quixel treatment this time around to add surface detail to the metal. (Since they all use the same material and textures, and they’re everywhere, I put more effort into making them look good.) Then there’s the various NASA cargo bags, which I made sure to place all over the more lit-up areas, in different states of being worked on or just stowed toward the side. As I mentioned earlier, they break up the uniformity enough to make everything look a bit more believable, and tell more of a story than the walls and floor alone.

Dealing with the number of props was easier when I used grouping, especially for the handrails, and something I’ll be looking into in the future is making simple Blueprints of frequently used environment pieces with props attached. From the moment I decided this was going to be a scene, I planned to have the handrails and bags, and the rest emerged after most of the environment was complete (folding tables, cables). These were suggested by friends I showed my work to, when I asked them what would make the scene look more lived-in and functional.

Finally, in the above shot there’s a big computer monitor and a flask, both of which had an unjustifiable amount of work put into them. I had planned to make similar personal items to place in the scene, but had already made these two separately from the project so I decided to dig them out and use them here. If I had made them to this level of quality only to be used twice in the scene, the effort would have been better spent elsewhere!

What this taught me

There are some big lessons I’m glad to have learned and have cemented into my mind from this project. First and foremost, I learned how to have a plan and be work-focused, not goal-focused. Beforehand, my biggest motivation-killer was having a solid idea in my head, sitting down to start work, and realising I would not be able to achieve it perfectly. It isn’t about perfect, it’s about more. Making more content means you get more practice, which makes a finished scene and a better artist by the end of it. Being more ambitious isn’t anywhere as difficult if you focus on each unit of work that goes in, instead of focusing on the end goal and trying to manually inject work to push yourself there. I hope that makes sense. What I do know is that many people I’ve talked to seem to have always known this intrinsically, and I would envy them for it.

Secondly, I feel like I’ve become a lot better at making something convincing that didn’t first exist in reality. When not working directly from reference, I’ve always had trouble thinking too much about what I’m making – being really concerned that it didn’t look functional enough, or that it wasn’t accurate enough to its real-life counterpart, instead of just going ahead and making something and coming up with a reason for its existence half-way through. Looking at a blank canvas with total creative freedom can be incredibly daunting, so just getting on with something–starting on a few vague ideas, then choosing a couple you want to keep, and working from there – seems like the most effective approach for a new project. For me, at least. I needed to give my projects more room to evolve naturally. There is no avoiding an ugly beginning in a lot of cases, but a rough start is also the fastest to get through hand iterate on, and if you want to change or remove stuff, it’s OK because you don’t love it yet.

Lastly, I learned Quixel. Trust me I’m not being paid to say any of this, it’s a really awesome tool and I highly recommend it. This is the first time I’ve ever been able to edit the albedo, normal, roughness, metallic, and occlusion textures all at the same time. It’s both empowering and a relief. I’ve heard Substance Painter is also great and is quite similar, so if Quixel’s reliance on Photoshop isn’t something you like, you can look into that; I haven’t tried it yet though.

Jack Parsons, 3d artist