Artist from Riot Games discusses his most recent project.
3D artist Sean Marino, who has contributed art to games like Spec Ops: The Line, Borderlands 2, Fuse, and Call of Duty: Advanced Warfare, discussed the production process behind his most recent R2D2 scene. Sean talked about modeling the environment and our favorite hero from Star Wars, building materials, working with the newest Marmoset Toolbag 3 and other details.
Introduction
My name is Sean Marino. I have worked primarily as a hard surface artist in the games industry for 5 years now, contributing to various titles such as Spec Ops: The Line, Borderlands 2, Fuse, and Call of Duty: Advanced Warfare. I currently work at Riot Games as a 3d Artist.
R2D2 Project
Thank you, I had such a great time working on it! When the idea for this project came about, I was fully immersed in Star Wars. Taking a quick look at the timestamps on my files, this was back in November of 2015. At the time I was reading the “Heir to the Jedi” book, actively watching the show on Disney XD, Star Wars: Rebels, the video game Star Wars: Battlefront was a few weeks from being released, and The Force Awakens was only a month away. Although I didn’t actually start working on this project until a few weeks after the fact, I was fully preparing myself for it by diving head first into everything Star Wars. I knew I wanted to make something from Star Wars, but I didn’t know what it would be exactly. I was scrolling through a bunch of reference I had gathered from starwars.com, from the “Rebels” show. For every episode of Star Wars: Rebels (as well as the show before it, Star Wars: The Clone Wars), you can find official concept art that was used for each episode. I had gathered up A LOT of it, and when taking a look through the files one day, I stopped to look at the sheet for R2. So much attention to detail is given to this character, and I really loved that, so that’s what I decided to make. Initially I had only planned on making a high-res asset, which is why there was such a huge gap from start to finish. Ultimately I think about 6 or 7 months had gone by from when I finished the high-res model, to when I decided to do the low-res, texture, and corridor.
The Corridor
The corridor was actually the last thing to happen in this project. I thought it would be a nice touch to put this character in an environment, instead of sticking with the typical neutral background for presentation. I also thought it would be a fun challenge, considering I tend not to do environment pieces like this very often. I did pretty much all of the modeling for the environment in 3ds Max. I knew I would be able to easily utilize the spline tools that max has in order to get exactly what I wanted, and keep everything nondestructive (very important). My typical workflow involves using lots of booleans, and I love the flexibility that max’s boolean system offers for iteration. Taking a look at the main parts for this corridor, I knew I’d only need a few basic shapes to get started: A door, a frame, a wall, and pillar.
For reference, I looked at shots from the movie, model kits, other fan art interpretations, etc. Once I had a basic understanding of the shapes, I began blocking everything out with splines, making sure to keep my stack in tact so I can always go back and edit.
The most challenging part about making the hallway was actually figuring out what I was making. A lot of the actual original set pieces were built from scraps, or just random parts, or at least meant to look like they were, so there seemed to be a lot of ambiguity in terms of set design. I had to force myself to get in the mindset of “you’re not actually making some vent or button or computer piece, you’re just making decorative shapes to hang on the walls.” None of the set really had any functionality, so I had to treat it as flavor only, which was pretty challenging since I tend to really enjoy making things that can work mechanically.
The scene itself though was fairly simple to set up. I used basically 3 main modular pieces to assemble it together.
R2D2
I still have this slightly weird workflow where I start all my models in Maya, and then move over to 3ds Max later. I think the navigation, movement/placement of objects, and general ease of workflow just goes more smoothly, once I feel like I have all my pieces in place, then I move to 3ds Max to start making use of the modifier stack. When I started modeling, I knew I was going to encounter a problem that a lot of people run into when dealing with cylindrical objects. You tend to run into the issue where when adding supporting edges, it creates pinching along the surface or around any inset you might make. So right away I had it in my mind that I was going to have to work higher res than I was used to. Traditionally I would use a minimal amount of geometry for my model, and let smoothing operations take care of the rest, but in this case, I started out with a much higher resolution base object.
image-1
After getting some blueprints of R2 online, I threw them into the scene and tried to best figure out what amount of sides would be easy for me to work with. I tend to care a little too much about technical details, but it just puts my mind at ease knowing these things ahead of time, so I can plan. So for R2 I started with a 320 sided cylinder. Through a little bit of trial and error, that number just worked out to be pretty perfect for what I needed to get started, knowing basically that all of the details on the body would more or less have the same poly distribution. To build some of the inset panels, I would just make one, copy it, rotate the appropriate amount to fit in the new spot (1.125 degrees per poly), and fill in the gaps. For that, obviously using a 360 sided cylinder would’ve been easier, but the amount of sides per panel would’ve been a little harder to deal with to get the right proportions.
Knowing that all I basically had to do was select groups of polygons, and cut them out, I just started by making selection sets or material IDs to mark my way through. Once I had all the pieces, I then did a similar approach in 3ds max as with the environment, by using instanced modifiers to make the highres. It was mostly “Shell” and “QuadChamfer” that did the trick for me. Adding that onto the basic shapes I had got me a good bit of the way.
r2stack
Throughout the entire process, I was making sure to pay as close attention as humanly possible to the details in the model, trying to make sure everything was exactly where it should be. Funny enough, I misaligned some things while rotating pieces to fit into their proper places, so the front-back parts are not entirely symmetrical, they’re actually 1.125 degrees off from each other!
Texturing
The approach I take to texturing tends to be the same regardless of the package I’m using. I start out with masking my base materials, then layering weathering effects on top. I try to nail down my materials with color only as best as I can before doing any sort of detail.
The benefit I’ve found recently of using Substance Painter here, was that I could create a lot of my color masks by selecting the geometry or UVs, so I didn’t necessarily have to worry about baking an ID map. And speaking of which, the ease of playing with masks and materials in substance has opened up the opportunity to create variations of this droid, for possible future projects as well.
For the specific texture details, I love that Substance has tons of generators from which to get a base started, but I tend to make use of a lot of alpha or texture maps that I’ve collected over the years. Some I’ll overlay over the entire texture for quick, seemingly random, noise (huge thanks and shout out to David Gruwier for his high resolution smudge textures).
I used Quixel Suite more so for the environment than R2, mostly utilizing nDo to add in normal details that I didn’t bother to model. I actually went back and forth quite a bit on the idea of doing all of the normal map details for R2 in nDo, but in the end I wanted to have a nice high-res model to bake from.
Since PBR became the standard, I find that some of the older methods of texturing (adding AO to your diffuse, adding a curvature map to your roughness, etc), still work really well! Even though it isn’t physically correct, if it looks good, I’m happy with it. Texturing is still in my eyes what can make or break a model. I remember hearing once that a good texture can make an ok model look good, but a poor texture will easily make a good model look bad, and so I’ve always tried to pay as much attention to texturing as I can.
With this project, it was actually my first time using Substance Designer as well. Although it is a tiny detail, the little lights that are on Artoo’s head were done using the tile generator, and blending colors in designer. The node graph itself wasn’t anything spectacular, but making it was a great intro to the program.
The Benefits of Using Marmoset Toolbag 3
Funny enough, I only got my hands on Toolbag 3 only a few days before finishing this project. Up until the beta went public, I was evaluating my work in Toolbag 2. I had been talking with a friend who was in the closed beta, asking if it really was that different, and then less than 24 hours later, it was in open beta and I was able to try it out. I’ve always loved Toolbag, from the very first one, and it gets better and better with every update. If I were to have done this scene with Toolbag 2, I think I might have been forced to use the indoor florescent HDR map to light the scene, which I’m not a huge fan of, so once I had access to TB3, the GI made use of the emissive planes I had placed in the ceiling of the hallway, and with the diffuse reflection, gave me the bright white light look that I was going for, and allowed me to use whichever HDR I wanted.
image-3
Due to the time frame that I got my hands on this in relation to where I was with this project, I unfortunately didn’t get the opportunity to try out the baker. I am eager to play around with it though once I move on to my next project. Animation is definitely something I’ll be looking to do in the future. I actually had considered before this release doing a lot of my presentation work in Unreal 4, since I could have access to things like animation graphs and such, but now I feel like I can easily achieve the same desired results with TB3. So far I’m really loving this update, even small improvements like the UI, or selection tools for your camera just make working with the tool so much easier.
Conclusion
What I found to be the most crucial aspect of recreating this scene, was just in the attention to detail. I was drawing from ideas that came from some incredible masterminds and visionaries, and I felt like studying and nitpicking every detail was the only way to truly do the originals justice. It can be very easy to ensure something has the general “look and feel” from something in the Star Wars universe, but these characters and worlds have been seen, studied, and loved by so many people, time and time again, throughout the years, and the moment any detail is out of place, or there’s something wrong, someone is going to notice. For anyone working on their own environment, or character, prop, vehicle or whatever, just gather absolutely as much reference as possible. While I’m working I’m not just looking at images from the movie, but also at books, toys, fan models, convention set pieces, original blueprints (if I can find them), anything I can possibly find in order to figure out what belongs. The more reference, the better!
Thanks again so much for giving me the opportunity to share my thoughts and process with you.
May the Force be with you!