Co-founder of Cubebrush Marc Brunet gave a nice interview about his character creation workflow.
We’re incredibly honored to feature an interview with Marc Brunet – an awesome 3d artist and one of the founders of Cubebrush. He was kind enough to share some information from his newest tutorial ‘Stylized PBR Character for Games‘. It’s a great look into the production, including a brief look at some of the most useful tools during the production process.
Hi everyone, I’m Marc Brunet the founder of Cubebrush, tutorial maker and ex game industry 3D artist! Nowadays, I’m mostly in charge of overseeing the development of new features on the Cubebrush.co platform but whenever I can find the time, I love to create various 2D/3D tutorials for artists.
Prior to working now full-time on my startup, I was a senior character artist over at Blizzard Entertainment for 7.5 years, and worked at Behaviour Interactive located in Montreal for a few years before that. While at Blizzard, I worked mostly on Titan (which as we all know was eventually reborn as Overwatch), then went on to do many heroes/skins on Heroes of the Storm and did some character work for StarCraft 2: Legacy of the Void.
Future of 3D Art
Honestly, I don’t think 3D scanning will ever impact much of game development. The beauty of game art is that it’s mostly fictional, and you can’t scan what doesn’t exist! There might be a few cases where scanning becomes an interesting option like on Battlefield or Call of Duty type games among others where the subject and art direction is more photo-realistic, but I’d have to guess it would only affect a very small fraction of the industry.
Stylized Characters Creation
It’s always a very interesting process. Just like working on a photo-real character, you need rules to go by and guidelines to follow. I’ll always start by asking myself what level of detail is best for the character, how to take details and simplify them to match my target, what kind of features I should focus on and what can be kept to a minimum. The more consistently I follow those guidelines, the better the result. Those decisions also heavily impact the tools and pipeline I’ll use, it varies for every project.
Generally speaking I’ll start from a base mesh that I have (built a few great ones over the years) but in this particular case I decided to start from only a ZSphere.
I start with the head, then extrude the rest using simple tools like masking and moving. I’ll normally get to a point where the character is all fleshed out with body fitting outfit, then I’ll decimate it, bring it in Maya and work on blocking out the rest of the bigger pieces. I’m just faster in Maya for that kind of stuff, but it can easily be done in ZBrush or your modeling software of choice.
Once I’m happy with the basic volumes, it’s back to ZBrush where I’ll do a detail pass on everything, starting from the focal point (the head) and moving down the body until it’s done. I don’t usually use much references but depending on the level of details, I might Google a few to make sure things look right. For this particular project, that wasn’t needed.
This goes back to what I was saying above, where I’ll set myself some guidelines prior to starting. Where the character will be used is obviously really important and it’ll drive a lot of the design/modeling decisions. If I wanted my character in a strategy game, I’d focus the design on bigger details, iconic silhouettes that can be identified at a glance. For something like a stylized action game or FPS where the character likely has a specific class or ability set, it’s important to focus on displaying that first. There are some pretty well-established classes for those kind of games and I wanted to make sure the character fits well within one of those. You can think of healers, tanks, snipers/assassins and the more regular soldier type, which my character is. Each of those classes come with their silhouettes stereotypes and it’s always good to roll with that as a base, then deviate if needed. It lets everybody understand what your character is about right away.
The materials I use are always a combination of metals, plastics, leathers and fabrics. Depending on the type or style for the character, I’ll shift the ratio of those types of surfaces. If I’m working on a healer type of character for example, I’ll focus on the fabric and leather materials, but will always try to include the other 2 as well, just in smaller ratios. It all has to do with the kinds of abilities and mobility the character needs to have. For a tank type, I’d flip it over and focus on metals and plastics. It’s important to have context for your character, then decide what is the best fit.
Here, I wanted him to be agile and quick like a marine infantry, so I made sure he had plenty of leather and fabrics where things need to move around or bend, and kept the metal plating for certain areas only as basic protection.
I used Substance Painter to brainstorm the final look since it always involved a lot of trial and error to really get the look you want, even if you had a solid plan going in. When it comes to color however, I can generally get a feel by looking at the color map in Photoshop after roughly coloring my bakes using gradient maps. I always aim for a ratio of 30% saturated/bright colors, and 70% desaturated, darker colors. Looking at the UV sheet in isolation makes it really easy to know whether or not I’m on target. If I had a second rule, it would be to use complementary colors. If my 30% saturated part is a warm color, I’ll try to have the rest be a colder color. You can clearly see it on my character. To offset the warm red armor pieces, I have greenish-blue leathers and fabrics.
I have pretty strong preferences when it comes to hair – I like the clay look, as if it was sculpted. That’s also the style I’ve used throughout pretty much my entire career so I know that it deforms well and it’s easy for animators to pour a lot of personality into it. I’ve never been a fan of hair blades simply because it tends to be used for more realistic styles and it’s not really something I go for. If anything, I think the best possible case is to have a stylized hair sculpt as a base and add on just a few hair blades here and there to improve the silhouette and hide potential seams.
I learned Marvelous Designer a few years ago and it can really do miracles for characters that need that extra bit of realism, but I never really used it since. I always like to sculpt the folds from scratch since it really allows me to do anything. One thing I do is look at real clothes and then remove 2/3 of the details, merging smaller folds into bigger ones and simply flattening out surfaces that are too noisy. There’s often a lot of trial and error involved trying to get the most aesthetic folds possible while still keeping in mind there needs to be symmetry as many places as possible without it being noticed.
I simply start from a naked base-mesh, then mask and extrude new subtools to use as the base for the clothes. If I need to create anything that is not form-fitting and needs a little planning, I’d start in Maya instead just like I did for the armor pieces, then bring it back in ZBrush for sculpting. Simple!
My main reason for using Marmoset is that it’s so easy to render turntables in it and export the scene to portfolio websites like ArtStation. Beyond that, their tools and post-process settings are probably the easiest to use out there to get exactly the look and feel you want. After plugging all my texture maps in and making sure everything looks the way it should, I’ll tweak the intensity of the gloss and emissive if needed, then move on to lighting. As far as lighting goes, even with image-based lighting materials, it’s important to have a good 3-point light setup to really make the character shine.
I start with the key light, which often represents the sun or a light closer to the character (brighter compared to the others) in the scene. For that one, a very light yellow or pure white generally works best. Then you’ll need the bounce light to go along with it. That light is pretty faint, and is generally the color of the ground, whatever it is. Finally, I always add a much brighter blue-tinted light to create a nice rim light on the character to help make it pop from the background.
The final steps consist of adding a bit of flare effect to soften the contrast, adjusting the saturation and exposure if needed and adding some sharpening to make everything crisp.
Adapting the Model to the Actual Game
Going into it, I had a set poly count of 15k in mind so I was always careful not to get carried away while I was doing the re-topology. This is a good general number for current games where the character occupies a good portion of the screen. Once the topo is done, I’m generally pretty close to my target from experience but in this case I still needed to shave off a few hundred polys which I did by removing all back-faces not seen when the character deforms, reducing the poly-count on areas away from the focal point like the feet and legs and getting rid of some more on rigid areas that don’t need to deform. The poly density should be pretty even throughout the character except on important areas such as the hands and face.
Then, when doing the UVs, it’s important to always use symmetry whenever possible to gain the extra texture space. For this project in particular, I ended up mirroring part of the hair, the neck and the ears for the character simply to gain that extra bit of space. The final UVs are always so tight each pixel is important. I could’ve easily decided to not mirror those, the majority of artists don’t, but that’s ultimately a waste of space.
Always keep in mind the textures will be heavily compressed and reduced in size in the actual game, so you really oath to use the space well. Even when working on portfolio pieces, I always recommend approaching the task the same – with a clear goal and budget in mind.
Marc Brunet, Cubebrush
Interview conducted by Kirill Tokarev.