3d character artist Gavin Goulden explains, why character production is so important for video games.
3d character Gavin Goulden explains, why character production is so important for video games.
My name is Gavin Goulden, I am the Lead Character Artist at Insomniac Games and am currently working on the upcoming Spider-Man game coming to the PS4. Prior to this, I was the lead on Sunset Overdrive, Bioshock Infinite, and worked on titles such as Dead rising 2, The Bigs 2, Damnation, FEAR 2, Dragon Age Origins, etc.
I was born in Canada and always had a passion for games, really more than art itself, I was attracted to comics, video games, and action figures. Really just mimicking a lot of what I saw in front of me, I wanted to create my own heroes and monsters then began drawing fan art (specifically of DOOM, XCOM, Venom, and X-Force – really anything involving monsters and muscle bound anti-heroes.) Eventually, I went to college to learn the technical aspects of creating 2d/3d digital art but eventually dropped out to work on my own material and develop a character art portfolio. After a year or two of learning on my own and improving my portfolio, I began working on a few mobile games and, eventually left to join a company working on 3d shooters (Piranha Games.)
After a few years there, I needed a change and decided to take up a few freelance offers I had and work independently full time. Getting the opportunity to work with clients such as Monolith, Bioware, Harmonix, and the military helped form a lot of better practices and allowed me to learn a lot of new techniques as well how to properly manage my time as well as promoting my own work. From there, I really was missing working with a team (working alone in your apartment for over a year can give you cabin fever), I joined Capcom Vancouver to work on their sports titles and eventually Dead Rising 2.
With work wrapping up on Dead Rising 2, and the games industry in Vancouver beginning to decline, I joined Irrational Games to be their character lead on Bioshock Infinite. This was an insanely formative project for me as I learned a lot about the creative process from some of the best in the industry, the team were all incredibly talented and dedicated to making the best game possible. This totally changed my mind on making art for games, and taught a lot of valuable lessons about creating assets for a purpose, to tell a story, what story that individual asset tells, and how everything fits into the world.
Irrational eventually changed directions, and I wanted to continue leading a team – especially with the lessons that I had learned – which lead me to an opportunity at Insomniac. I loved the style of Sunset Overdrive and the unique take on “end times” spoke a lot to me as far as art direction and attitude. From there, things have gotten even better as I work on the next title focused on a childhood hero of mine, Spider-Man.
Building Game Characters
There are a lot of different considerations you need to take when creating a game character versus creating just a sculpt in Zbrush. When creating a digital sculpt that will only live in your portfolio as beauty shots, there are virtually no limits to what you can do. You can create to a camera angle, not worry about topology or vertex count, and most likely rely on some post processing after you render that sculpt out. For games, it’s totally different. On a technical level, you need to consider the scale of the character, how sculpted information will bake down onto a texture, the posing of the character and how it will deform once it has been converted to a game mesh, and how a character will read once in the game camera versus the Zbrush viewport.
Beyond that, a major difference lies in the purpose of the content you are creating. For a sculpt, it can essentially live on it’s own as a singular asset. While the story it tells can play a big part in making it an appealing piece of art, creating a character (or really any asset) for a game can be way more involved. Everything you create needs to live in the same universe, follow the same art direction / style, and be unified in quality. Then, this asset is going to be viewed from a totally different perspective and often needs to serve a clear purpose to the player. Certain silhouettes, color schemes, sizes, etc. Speak a language to the player and conveys the purpose of the asset. Besides the gameplay purpose of the art, the answers to questions you ask when creating needs to extend further than “Because it’s cool.” What does this character do for the player? How does it progress the plot further? Is it a friend or foe? Personally, I think just sculpting a character has a lot more wiggle room when it comes to this and you can generally focus on the singular asset versus how it acts a piece to the bigger picture.
It is definitely important to consider how the character will move and deform in the game. The basic pose of a character, the angle that their limbs are positioned and how the points of articulation are defined will greatly effect the outcome of cleaner joint deformations during animation. Arms that are too far down will look good in resting poses but fall apart during extreme ones, legs that are too close together will lead to complications during baking and skinning, hands that are in too extreme of a pose will be difficult to rig, etc. Generally, you want to have all joints in a neutral position that will allow for movement in both directions. When you create the low resolution version of your model, there will be edge loops lining the joints that squash and pull the faces in between them. The more compressed these loops become, the more distorted the geometry will become once the underlying bones are moving in the opposite direction.
Different Character Tds have different preferences, like a preferred bind pose, bone angle, and things like that, but the principles are all the same and what applies best to your current project. Generally, there is going to be a set skeleton throughout the project which most are derived from in order to share animations and make the rigging process smoother.
My personal workflow has become a lot more organic over the years as I trust my work more and am more comfortable with things looking bad (sometimes really bad) before they look good. Originally, I followed a very linear path of creating a loose concept, then base model in Max, then sculpting in Zbrush, then low poly, bakes, textures. Now, I bounce back and fourth a lot. I’ll usually begin right in Zbrush, will import abase model if it’s part of a bigger project, and just dive right into sculpting up details and keep it really rough. At this stage, I’m focusing on bigger shapes and how this character will read in game, mostly just focusing on the main elements of the character. I usually don’t even do a traditional concept and consider this phase my “concept stages” where I can change the character up more without causing much damage and really just let the character speak to me during the process. Once I land on something I am happy with, I either pull pieces from this sketch or create a cleaner version of the model in Zbrush or Max, then continue on sculpting.
The whole time, I’m collecting references to inspire what I’m working on. Unless it’s at work, I usually don’t follow a strict concept and like to keep it loose. Generally, before the project even starts, I will collect references and store them locally or have a page on Pinterest. This will usually consist of a few different key elements I know I want, like types of armor, how clothing generally is constructed or pictures of a garment I have in mind, or even inspiration for color usage and rendering styles.
One of my favorite parts of creating character art is texture painting. While you can get lost for days adding tons of detail to a character sculpt, painting really brings it to life. This process has changed greatly over the years from a manual effort in Photoshop – creating every detail by hand or using photo sourced information, to a texturing specific application like DDO or Substance Painter.
The industry standard is becoming applications such as this mostly due to the adoption of physically based rendering which requires consistency throughout the project for specular, gloss, and albedo values. Beyond the much needed consistency in a project using PBR, the process of painting materials in an application (defining metals, cloths, wood, etc.) versus manually creating these materials in individual textures is a huge time saver.
My process basically involves creating a material ID map on my high resolution model through polypaint in ZBrush (for example: red equals skin, yellow equals gold, green equals leather, etc.) and bake that information out as a color map to be used in Substance. Taking this ID map, I assign the basic materials which I had defined earlier and begin working in layers of detail. Considering the story this character has to tell can determine which type of information gets added. What environment the character belongs in, the character’s age, if they have seen battle, etc. Based on this information, extra detail can be as simple as a bunch of scratches, or a very involved layer of damage and rust.
I’ve always had an interest in costume design, and the type of clothing a character wears can greatly change how appealing they are to the viewer. Dressing your character in a certain way is very important as the choice you make for the garment reflects the type of choice your fictional character would make. They exist in a world and, although completely made up, wake up every morning and make a choice about what to throw on their bodies, and the range of possibilities are endless. A more shady character can conceal their identity, a proud noble has clothing made of expensive materials, a brute has battle worn armor, and a post apocalyptic survivor wears whatever they can find – even items that aren’t normally clothing. This enhances the appeal of the character as it tells a story within itself, it ties the character into the world, and generally shows that the character itself has a thought process – it gives more purpose to the character when their clothing is rationally thought out and plays into their function within the game.
When creating a character, I like to play with a few different options. Much like scan data, this comes down to the need of the project. You can use an application like Marvelous Designer, which can help generate realistic clothing fairly quickly, or sculpt it out in Zbrush. In my opinion, there really isn’t a better way of the two as it depend son the type of garment you need to create. Something really involved or not based in reality can lead to some pretty frustrating times in MD as you’re expecting non-realistic results. Getting really accurate drapery in Zbrush can be time consuming and could be better suited for the simulation power in MD.
Regardless, the best tool of them all when choosing garments for your character, is Reference. Find tons of reference and understand what type of clothing your character would wear, how it would be constructed, and how it would react to gravity. No matter how realistic or stylized, or if it’s made by sculpting or simulation, the garment needs to make sense for the character, and the best way to determine that is by using reference.
Personally, I feel that there is a place for both photogrammetry and stylized art and it really comes down to the type of project you are working on. If the goal is a photo realistic game with nuanced information in your models, photogrammetry is a great tool to have as it captures all of that data and saves a ton of time trying to replicate what you can, basically, take a photo of and have it as a workable asset. Not only that, photogrammetry plays a huge part in many higher end performance capture systems and really is a key element to having life-like animations in a cinematic setting. However, there is almost always a need for the human touch to give this information style to fit inside the game world and not everything can just be scanned in. Unless the game is trying to mimic the real world verbatim, artists will need to add clothing, armor, detail, distort proportions, etc. To make it appealing.
There are also many stylized games out there that are just as profitable as, what would be considered, a photo-realistic game. There will always be a need for escapism, for fantastic worlds, and the creation of things that simply do not exist in our world. Some teams also simply can’t support the workflow involved with photogrammetry, the time and resources spent obtaining data and cleaning it up, possibly even the cost to support the full pipeline properly, may be out of the question for independent projects. Plus, to be totally honest, I think a lot of artists started by wanting to be creators, not technicians. While the job certainly has a large technical side and different projects have different needs that may not line up perfectly with the individual’s personal choice, no amount of scanned data will remove the desire to create new things.
While photogrammetry is and will continue to be a powerful tool used during development, especially for projects seeking a realistic rendering style, there will always be a need for good sculpts and creating things that do not exist in our world. I think the healthy way to look at photogrammetry is to see it as a supplement, like with scanned models and materials, this information can be fused with creations to help speed up the process, make the final results consistent, and generally remove tedious tasks while allowing artists to focus on other areas.