Weapon Production: Building, Texturing, Lighting

Patrick Sutton shared his thoughts on the weapon production: building, texturing, and lighting workflows, weapons for games and what they sometimes lack.

Introduction

Hi there! I’m Patrick Sutton, and I work at 343 Industries as an Environment Artist. I shipped Halo 5 at my current job, and before that, I was an outsourcer at Liquid Development, where I worked extensively on Firefall and contributed to a few other projects including Halo 4 and Wildstar.

I’ve been modeling for 18 years, 8 of them professionally. I started out making levels and guns for Half-Life and continued modding more games as they came out, such as Morrowind and Fallout 3. I am entirely self-taught, as there weren’t a whole lot of resources out there for a 13-year-old at the time, and I didn’t realize there were online communities for people who did what I did!

Technical Skills

I do quite a bit of programming on the side. Usually, it’s just tools for my 3D package, but I used to make local multiplayer games as well. The primary benefit I get from programming are the skills of creative problem solving and thinking procedurally. The thought process is the same for things like shader networks and Unreal’s Blueprints. This helped me greatly when building an ocean foam shader that handles obstacles, like the one seen in the Halo: Infinite trailer.

I think tools like Unreal have really put great power into artists’ hands, as the systems they have built are designed to be discoverable and easily understood. A graph is much easier to understand and debug than lines of code for many people.

Substance Painter is another story entirely. Allegorithmic has made a pair of very cool tools with Substance Painter and Substance Designer. Both of them are designed around the concept of materials, which frees the user up to think about what really matters instead of juggling layers and texture sets in photoshop. Personally, Substance Painter is the best 3D painting application I’ve ever used.

Weapon Production

There are a few different ways to go about building guns. Depending on access to real weapons, you may be able to actually just measure things. I don’t have access to those, so I instead hunt down useful dimensions and use various measuring techniques to ensure the accuracy of my weapons. Since cylinders have a known diameter, you can use the barrel (for instance) to measure the width of something even from a perspective photograph.

Speaking of photographs, I try to accumulate as many photographs as I can. For most projects, this is anywhere from 100-300 photographs. For vehicles, I’ll have even more! Photographs are crucial to making an accurate model. Not only are you able to see how details are made, but you can use them to hone in on measurements that aren’t obvious from other angles.

I usually make the cartridge the weapon fires from dimensions found on Wikipedia, find out basic dimensions where possible (overall length, barrel length), and make educated guesses based on the relevant unit scale. For European weapons, this means the metric system, and so while modeling a pin or something I can guess the size in millimeters. This is harder with English and American weapons since they use the imperial system which has annoying fractions of non-decimal units.

Weapons for Games

My weapons aren’t actually suitable for game use (except maybe for something like Escape From Tarkov, which focuses heavily on the mechanisms inside of weapons) because I model all the internals and break up parts for better exploded displays. For most games, you will never see the insides so it’s important that you only build what the player can see, and plug holes that give visibility into unimportant parts of the inside of the weapon.

Triangle count will vary between studio to studio; I use 30,000 tris as an upper bound for a first-person weapon and aim for 20,000. This is a tricount I have seen at several studios, but I have also seen ones that go far higher and far lower. Lower triangle counts are sometimes for efficiency on multiple platforms, and sometimes because the weapon can have many attachments which themselves add a large number of triangles.

Readability is accomplished in much the same way as it is in any other design process. A strong hierarchy of forms, colors, and values is important to make the design easier to read to the human brain. I don’t focus on this a lot with my real-life firearms, but for my design work, I focus on it heavily. A good way to learn these principles is to study graphic design and typography. All of the same rules apply, but it’s easier to learn on a 2D plane.

Authenticity is accomplished by understanding your source material. If you are designing a weapon for a normal human body, it’s important to understand how weapons work, and why they are designed the way they are. This includes layout and ergonomics. If you design a weapon that is too big to be easily held, cannot be aimed correctly or has proportions out of line with its size/type it will feel wrong.

I really like what the futuristic Call of Duty weapons have been doing for authenticity. They feel believable, even if they have bonkers functionality. One example that I really like is the laser rifle that can be split into two pistols from Infinite Warfare.

The only animator I have ever worked with on firearms is myself, so I can’t really speak to production stuff like that.

Texturing Weapons

The first thing I do is to try to make weapons that have many materials on them so that I have a good basis for an interesting material breakup. I have a few models that I’ve finished but am hesitant to post because they’re simply black metal and black plastic – a Saiga-12K, a pistol-caliber AR, and a plastic AK74, for example. One I’m working on now is an Uzi that has black foregrips, a green handguard, parkerized receiver, and different metal colors on the top cover and magazine. I picked that model of a weapon specifically because it has many different materials.

One of the most important things to focus on in a PBR model is the roughness. I make sure that I have a variety of roughness values across my model. In the case where two materials have similar roughness values, I make sure to have different patterns in the roughness. For instance, a metal might have a bumpy look to it because of how it was cast or painted, while plastic might have a smoother appearance. If both are similar in roughness, I might make sure that there is less detail in the plastic than the metal to separate the two.

My Substance Painter workflow doesn’t actually involve Substances or scanned materials; I only use it as a 3D painting tool. I’ve made a set of overlays that I carefully apply to the model, typically only in the roughness, and I make use of generators to get edge wear, automatic dirt, and so on. When using generators it is critical that you make your own custom grime. This gives you complete control over the final look.

What Game Props Sometimes Lack

I haven’t seen this in professionally made artwork, but sometimes while playing game mods I notice that the modeler hasn’t accounted for what will be most viewed by the player. For a first person weapon, it’s important to spend your triangles and texels on stuff that is closest to the player’s face.

What is close to the face depends on the animations and use of the model – if the player will regularly see the model in a customization screen, even texel density makes sense. If they will only see it in the first person, and animations don’t move the end of the weapon towards the camera, it’s reasonable to decrease texel density there so that more texels can be used on parts closer to the camera.

Related to this is triangle usage. If your game has aim-down-sights functionality, it’s good to spend extra triangles on the rear sight. If it’s a round peep sight like on an AR-15, spend enough triangles that it doesn’t appear polygonal at all. The same goes for silhouette details. I have seen a lot of weapon barrels that are obviously polygonal which would benefit from doubling the number of triangles.

It’s worth pointing out that my weapon models are made for my own satisfaction, and to be viewed on ArtStation, so I tend to use an even texel density on everything except deep internal parts – the insides of grips, the insides of a dust cover or receiver, and so on. It’s important to make your work fit for purpose.

Rendering

I use Marmoset Toolbag 3 with default shaders to render my game-res models, and KeyShot or Blender to render my high-poly design work. In Toolbag, I have a series of cameras, which contain copies of the model and all lights needed for a shot. This makes it easy to manage variations on the model and make sure you can modify a shot without messing up any others. For instance, having a shot that has the bolt retracted, a shot with a folded stock, or a shot where the weapon is disassembled.

Lighting

The most important thing when presenting models is lighting. Great lighting can improve the appearance of a mediocre asset, and bad lighting can ruin an amazing asset. I try to light things so that the model pops off the screen, and so that the materials are made evident. I accomplish this in two ways.

The first is getting good rim lighting. I turn the environment lighting down to about 5% so that my model appears pitch black. Then, I place two rim lights and adjust them until I’m happy.

The next light I place is the key light. This is the dimmer than the rim lights, but brighter than the fill light, and is the one that I use most to show off materials. I spend a lot of time trying to get a specular response that shows off the form and the materials to the viewer.

After the key light, I place a fill light, to get some light into anywhere that’s still too dark. This should be your dimmest light. If your fill light is a significantly different color than your key light, you can probably make it as bright as the key light.

Finally, I might add a few more lights that enhance what the key light is doing, or highlight a point of interest. In this case, I thought it was looking a bit flat and added a bright light to the upper planes of the model.

I think this is a good workflow for any lighting type. I personally like to light things very hot, but there’s no reason that this workflow can’t be used for different styles.