Hard Surface Modeling Tips and Tricks

Introduction

Hi, my name is Trevor and I'm a 3D game artist currently residing in Los Angeles. I've been working in games for a while – my first project shipped in 2004 where I helped out as a character artist intern (Harry Potter and The Prisoner of Azkaban on PC). And I've been at it ever since. Hard surface modeling is my primary skill set, and character and creature work is my secondary skill set, which I mostly do for fun in my free time. I suppose if I was smart, I would pick one and focus on it, as our trade is getting quite specialized. But alas, I'm an artist, so I tend to follow my heart and not my brain sometimes.

As for how I wound up doing this for a living – it was basically just a random chance. College and higher education wasn't an option for me after high school, so I spent quite a few years just doing blue-collar work – swinging a hammer or a chainsaw or driving trucks and forklifts, etc. This was also pre widespread internet access, so the scope of my world view was a bit limited, and I wasn't too aware of all the possibilities out there. I always had dreams of going to art school someday, as art was always my primary interest and passion, but I wasn't exactly sure how I would make it happen or even what sort of art I would study. I figured I would either do tattoos or graphic design. I had no clue about computer graphics at the time. I actually thought the cool computer graphics I saw in movies were made by programmers and genius mathematicians, not actual artists.

But, as destiny would have it, I got laid off from a warehouse job I was working and found out I could go to school while I was on unemployment, basically getting some financial assistance from the state for education. So I started looking around at the local community colleges for art classes, and I stumbled on computer animation. I didn't really know anything about it, and I had never really even used a computer before, but I saw the curriculum involved drawing classes, so I signed up. This was around 1999, I think. And after a few years of lots of effort and Gnomon videos, I managed to create a whole new life for myself, which I still feel blessed about to this day.

The Most Important Bits of the Process

This is a good question. I think it's important to create and maintain a decent level of focus and clarity with your craft. It can be easy to constantly get sidetracked or distracted with new tools, techniques, or software, etc., which oftentimes can wind up becoming outdated or getting replaced in time. Whereas spending time working on fundamentals, like design, composition, drawing, sculpture, painting, anatomy, or more technical things, like understanding topology, how surface normals work, or how graphics engines and shaders work, etc. will serve you long term, no matter what the latest techniques and software are. Fundamentals tend not to get replaced or outdated. So I tend to be selective about how or where I focus my attention, trying to get the best long term return on my investment of time, and not getting too distracted by the latest shiny cool tool. And if a tool does look promising, I'll wait a few months for everyone to figure out all the bugs and quirks and post their solutions on a forum somewhere, so when I do jump in all the problems will have been mostly solved and posted online, and I can hit the ground running.

Working in MODO

I switched to MODO like 9 years ago basically because it was designed somewhat specifically for hard-surface/product modeling, and I was all about hard-surface modeling. I found it to be very nimble, easily customizable and it had a really good set of modeling and UV tools, etc. The result was a much more fluid modeling experience. And it was cost-effective, so I could have my own personal copy and show up on a job and start producing content on day 1, as opposed to relying on whatever software an employer would provide – switching back and forth between Max or Maya, getting caught up on using one or the other, figuring out which scripts work and don't work with whatever version of software I was provided, etc., which tends to be a bit of a time sink. That being said, I always output my models to the studio's application of choice, so other team members have proper access to whatever they might need with the asset.

The Modeling Approach

For me, when I'm modeling more complex or functional machinery and vehicles, it's all about the mass-out. I tend to create fairly resolved, but easily editable mass-outs - establishing the aesthetic component pretty well - shape, form, proportions, stance, attitude, composition, visual flow, and maybe basic material break-ups, as well establishing mechanical and technical elements like functionality, baseline/target topology/density, meeting design requirements and gameplay needs, seeing the asset in game in relation to other assets, making sure leads/managers/collaborators can see and understand the asset in its entirety before completion, and getting feedback, and testing and refining. I try to identify and solve as many questions and problems as possible in the mass-out phase because there is almost always some unknown or unforeseen variables or considerations on a more complex model that players interact with. So when it comes time to finalize the model, I can do so quickly and with a clear vision, as well as minimizing the likelihood of having to go back and make changes to the model after it's been completed. I find the more I have to rework a model after it's basically done, the more it sucks the life out of the model and the jankier it becomes. So I try to avoid that scenario as best I can. Also, the mass-out phase is a great opportunity to explore and brainstorm about potentially cool ideas or features that hadn't been considered up until that point, which can make for a cooler model or more compelling gameplay. In my experience, the design isn't done at the concept phase, it evolves as it transitions to 3D. Especially if it has a fair amount of functionality.

A general guideline I tend to follow is to be as “non-committal” with the model for as long as possible, allowing for more room for exploration, feedback, and adjustments. It's not uncommon to get design requests from designers or directors that aren't fully resolved yet, so it's sometimes up to the artist to help them find and clarify what they need or want, or how it will work, and integrate everything into a cohesive, functional, cool-looking asset.

MODO has some really cool features for hard-surface modeling, although I'm pretty sure these features are all more or less available in all packages now. One thing that makes life a whole lot easier for high-res/subd modeling is using Edge Weighting instead of Support Loops for creating the high-res geometry. Instead of running Edge Loops around geometry, I simply select the edges I want to add the bevel to and just push a button that determines the amount of bevel or roundness (10%, 20%, etc.). That's it. Sort of like Crease Edge in Maya, except it creates an actual bevel that I can control. So in some cases, to high res a piece of geometry, it is literally 2 button clicks. Just click “Select edge by xxx angle”, and then assign the edge weight. It’s not always that simple, sometimes I need to select a few more edges by hand, or add a few general support loops across larger spans, but that's it. So it's way faster than traditional Subd modeling. Also, it is completely editable and non-destructive, so if I do have to make changes to the high-res mesh, I don't have to redo edge loops, etc, which is another nice win. The drawback of this technique is that I sometimes subdivide the model up to 4 times, instead of the usual 2, so I can get the most control over edge bevels. So I do wind up with geometry that's a bit heavier. But with a decent computer and good scene organization, it's not a problem.

This edge weight feature also works well with MODO's “MeshFusion'”, which is a procedural subdivision boolean modeling tool, using geometry to combine or cut out shapes and forms, etc. This isn't something I use on vehicles – it's more so for more complex modeling, as a weapon scope, for example. MeshFusion is also non-destructive and perpetually editable, so between the Edge Weights and Meshfusion, you can create very complex high res models and forms very quickly and completely non destructively, with total freedom to change or update anything anytime very easily. It's a very robust and user-friendly technical modeling process. There is also a “light duty” variant of this process that is equally robust called MOP Booleans, which can be combined with a rounded edges shader and baked out, for ridiculously fast “high res” model creation.

Working on the Details

I tend to approach a model in 3 phases or passes – Primary forms and structures, secondary forms and structures, and then details. As for details – the higher fidelity, tertiary bits, and pieces, they play an important role in bringing a model to life. The details and greebley bits do a lot to add functionality, story, visual interest, and scale to a model. I think of details like seasoning on a meal – they can do a lot to enhance the flavor of a well-done model.

For details, I have a kitbash kit I've amassed over the years – nuts, bolts, rivets, latches, hinges, greebles, etc. Details are the final stage, the icing on the cake. Once the high-res model is basically done, I'll do the “nuts and bolts” pass. Although, my workflow has been evolving a bit lately. For both, high-res poly modeling and in Zbrush. I am starting to do less of a detail pass on the high-res models because it is so much easier and more flexible to do so in Painter. Again, going back to my idea of being as “non-committal” for as long as possible, leaving the details for Painter gives me more flexibility to explore and iterate. Also, not needing the higher fidelity detail on my models cuts down on the polycount/memory footprint a bit, which is always nice.

Working with Materials

One of the coolest evolutions of game content creation in the last few years has been PBR materials. It adds a whole new level of fidelity to a model. And now that we can have a variety of real-world materials to work with, my modeling has evolved a bit to accommodate this new tech. Specifically, I think about materials a lot earlier while I am modeling. I constantly think about what the surfaces will be or maybe and look for ideas or excuses to add a good amount of material variety or material break-up. I look at the different potential material/surface types as musical notes, to be used to create a compelling visual symphony for the viewer. These more dynamic materials are a great compositional tool to lead the viewer’s eye around a model, to accentuate focal areas, and to tell a story. Really, the texture is essentially an act of storytelling, a narrative about the object, and realistic materials can add a lot more dimension to a model's story.

For me, the normal map and the roughness map are the most important textures. Light moving across forms, and reflecting back off the forms is what makes a model seem dynamic, realistic, and compelling. A 3D model is essentially a sculpture – and the essence of great sculpture is shape, form, and light interplay – which is designing cohesive, appealing and aesthetically sound forms, shapes and shadows, and their respective relationships. Again, always keeping in mind the fundamentals. One thing I often do to facilitate nice forms is I will hand triangulate certain parts of a low-res/in-game model on it's more rounded forms to make sure faces are properly convex and/or/not concave so the surface is flowing nicely. This might not always be necessary, and I might not always do it, but it is a consideration, as this will help make the normal map bake and light interplay that much better.

Also, with hard surface models, the margin for error in regards to a perfect normal map bake is much smaller than organic models. Between a large glossy surface or the potential of a player taking cover behind a vehicle that winds up taking up 1/3 of the screen, flaws in a normal map bake can be very noticeable. The diffuse map is almost secondary – it's all about the light interplay on the model, not what color it is. And with hard-surface models, there is a lot of opportunities to really push surfacing – from glossy to matte, dielectric or conductive, clean to worn, scuffed and scratched, corroded, etc.

Testing

For vehicle work, it's generally pretty wise to get to testing early and seeing how things are taking shape, even if it's just a static prop. Initially, just to check the size and proportions and to see how it might relate to other items in the game/environment or the in-game camera FOV. Then, there are also gameplay considerations – for example, sometimes there may be certain height or size requirements or preferences for things like taking cover behind the vehicle, sightlines around, through or under the vehicle/asset, or animation needs, like player entry, exits, and interactions. So establishing how players will interact with the vehicle/asset in a variety of scenarios is something to keep in mind and resolve early. I always assume that there will be some sort of problem or consideration that pops up that we didn't think about, which will cause a bit of rework. So, I'm always trying to identify those issues before they sneak up on me.

Another thing to consider is the actual functionality of the vehicle, assuming it is not static. What sort of moving parts will it have? It's definitely a good idea to build out and resolve things like suspension, turrets, cockpits, landing gear, etc. Early on, and get a rig on them to make sure they are working as expected and looking good. This is again an opportunity to add or enhance the design. Often times, talking with the technical artists or riggers/animators, they can have additional suggestions or ideas about what is possible that I might be unaware of, or some idea they might want to try. Again – thinking ahead, and collaboration are keys to a good vehicle work.

This is also a good time to explore some other potential cool non-critical/aesthetic rigging/animation ideas – like spinning drive lines, mud flaps, fan blades, whipping antennae, FX, or other “non-necessary” functionality to add some secondary animations and additional realistic interest and life to the vehicle. These can often be pretty cheap – just a single bone with an animated spin, a basic physics object, or a simple FX emitter shooting sparks or dripping fluids – but can add a lot of life and immersion to the vehicle experience. Which is also another thing to remember. A fully functional vehicle is an experience. So always be thinking about how immersive the interaction will be or can be.

Advice on Building Games

I would break this down into 3 core domains – Aesthetics, Functionality, and Execution. How does it look, how does it work, and how is it implemented?

Aesthetics – This is the artistic, purely visual aspect of asset creation, where solid art fundamentals come into play. Understanding and implementing things like composition, proportion, silhouette, hierarchies of shape/color/value, visual weight, visual flow, structure, anatomy, gesture, stance and attitude, negative space, etc. These concepts are all tools to create visually appealing and immersive graphical content. These tools will draw a viewer in and keep them engaged, and if done well, will capture their attention, imagination, and curiosity.

Also, these are fairly universal concepts that apply across all mediums – landscapes, vehicles, weapons, characters, creatures, etc. “Gesture, Stance, and Attitude” isn't just for characters as one might assume. They apply to vehicles just as much, for example – does the vehicle look and feel like what it is supposed to be and do? Gesture, stance, and attitude can help sell vehicles the intended purpose and backstory, and in effect, it's personality - aggressive, tough, fast, stealthy, dangerous, benign, nimble, etc. “Composition” isn't just for illustrations. I think about the “composition” of a vehicle from multiple angles. Does it look cool/interesting/compelling at the front view, front 3qtr view, side view, rear 3qtr view, and rearview, from above and below? Does each of those angles have focal areas, areas of interest, nice visual flow, nice material structure/breakups, etc? “Negative Space” adds complexity to a model by breaking up its silhouette, adding dimensionality, and making it look more realistic and functional. I always look for opportunities to break up a model's silhouette, as well as add depth, with recessed areas/vents with mechanical bits/details to sell the idea that there is “stuff inside”.

This is where we come back to the idea of making sure to keep in mind and work on the fundamentals of art and design that are timeless and universal, no matter how the tools or techniques might change or evolve.

Functionality – It's easy to get caught up in a cool design and not fully consider how an asset will be integrated or function. Keeping in mind what we discussed earlier – always consider all the potential upstream/downstream variables and requirements for the asset beyond just making a cool looking model. If you create a good looking model without those considerations, and you have to go back and fix or redo your model, your model will most likely look worse. In my experience, very rarely does going back and redoing or reworking a model to fix something improve it aesthetically. It usually makes it a little bit worse. Also, the rework tends to be a time sink – so if you are not thinking and planning ahead, you can wind up spending more time than necessary doing artwork that isn't as good as it could have been. Make it a professional goal to avoid this scenario as much as possible. You may not be able to completely avoid such scenarios, but I have found that with some forethought and effective communication, you can substantially minimize it.

Execution – Execution is essential “Craftsmanship”. A thoughtfully built game asset that takes into consideration efficiency, optimization, and organization to achieve the best looking and performing results, as economically as possible. This includes:

Resourceful and economical usage and/or creation of materials and textures – Materials and textures take up a lot of resources. Generally, it can be good practice to try to use the least amount of materials and textures you can to get the best result. You can get a lot of mileage out of minimal textures and materials if you are creative and resourceful enough. Also, different game engines handle this data differently, so ask your tech art or programmers what the biggest graphical bottlenecks are – is it texture sizes, or texture amounts, or amounts of materials or types of materials? Find out what your engine likes and dislikes the most, and then tailor your process accordingly to maximize awesomeness.

Lean and organized topology – Generally speaking, every vertex and edge should serve a purpose. If you can remove an edge or vert and the model's shape or it's functionality doesn't change, then it is potentially unnecessary. This is a generalization, as there are exceptions - a geometry that is needed for deformation/destruction, or geometry that will have a different material assigned, or perhaps for UV purposes. Whatever the case, an edge or a vert should be serving a specific purpose. Also, avoid ngons or leaving overly convex/concave poly faces to be arbitrarily triangulated by software at some point in the process, potentially causing baking and/or shading errors and surface lighting anomalies. Conscientiously craft your geometry for the best results. Like I mentioned before, sometimes hand triangulating parts of geometry, so the light turns on and across the forms smoothly. Also, adding a little extra geometry in highly visible rounded areas, like a arc of a wheel well or the arc of a vehicles roof, to help create rounded and smooth forms - as it will create the illusion or imply that the rest of the model is also just as high fidelity/smooth and round, even if it's not. If you spend your polygons wisely, you can afford to use a little extra geometry in places like this to create a better looking or more convincing model.

Well done/packed UVs – Straighten and square off UV islands to maximize space/pack and keep edge bevels in the normal map crisp (since edges are often where UV seams are), especially if/when textures get downsized, as they often do, and mirroring as much as possible. And not just mirroring one side to the other – for example, perhaps you can mirror a vehicle's driver side front suspension and wheel well to the driver's side rear, and then mirror the driver side suspension and wheel wells over to the passenger side. That's a 4:1 optimization, as opposed to a 2:1 optimization just mirroring one side to the other. That might not seem like much, but it all adds up, and that's the sort of efficiency and resourcefulness that can add a little extra fidelity to your asset that can help make your work stand out from the rest. I'm always looking for opportunities like this to squeeze every last drop of awesome out of a model, and you will find them often if you are looking.

Proper Baking – Understanding tangent spaces, surface normals, triangulation, sound topology, and UV set up is crucial to do the work nicely. Spending a little bit of quality time at Polycount or wherever and really getting to understand these technical elements will make life much easier and more awesome. The more you can eliminate guesswork and trial and error when it comes to baking, the more you can focus on quality art and meeting deadlines. And like I mentioned before, the margin of error for hard-surface baking can be pretty small, so getting dialed in on the technicals of baking will save you a good amount of time and make your work that much better. There are some great sticky threads at Polycount by Joe “Earthquake” Wilson that pretty much cover everything one needs to know on the subject. So, no excuses!

I hope I may have helped or inspired a few folks out there and added some value to whoever read all this. Feel free to hit me up with any questions or comments or what you have.

Trevor Crandall, 3D Artist

Interview conducted by Kirill Tokarev