AI applications are only beginning to materialize. It's so exciting what neural networks will do to art workflows in the coming years.
W T F!!! Why, really don't understand, EA = just a fucking looser company!
Amazing... Congratulations for the new way to show information.. I hope they could use this to teaching.
Tim Simpson did a detailed talk about the way he created this wonderful sci-fi installation, relying on a limited number of assets, and clever UE4 tricks. Also be sure to check out a little project Tim is working on – Polygon Academy.
My name is Tim Simpson, some of your audience might know me as Pixelmasher from Polycount. I’ve been an environment artist in the game industry for the last 10 years or so. I had a chance to dabble with 3DS Max in high school and then spent a few years learning on my own as a self-taught artist. During that time I was posting my art on various forums for feedback. One day I saw I had a private message in my inbox from the art lead at a small studio in Vancouver who had seen my work and wanted to bring me in for an interview/art test. I managed to jump through all those hoops and got my 1st industry job working on SOCOM: Confrontation for PS3.
Since then I have worked at a handful of studios in Vancouver such as Relic and United Front Games, on Warhammer 40k: SpaceMarine and Sleeping Dogs. I spent the last 5 years at Ubisoft Montreal working on Splinter Cell: Blacklist, Farcry 4, Watch_Dogs 2 and a currently unannounced title.
I recently left Ubisoft to go “all in” as a creative entrepreneur and am in the process of building my own company, Polygon Academy. I’m currently creating a ton of content for game artists and students. Everything from art tutorials to general industry advice based on my experiences over the last 10 years. So far it’s been an interesting change of pace, while still allowing me to create game art on a regular basis. The goal is to create the resources I wish I had when I was struggling to break into the industry all those years ago.
Data Core Project
Having spent the last 5+ years working on mainly photoreal/modern day subject matter I wanted to do something drastically different for my first personal project in a while. I have always been a fan of sci-fi art but definitely felt my skill set in that area was a bit lacking. So the goal was to attack one of my weaknesses and learn something new in the process.
Originally I was inspired by this piece by beeple. and thought it would be a fun scene that I could pump out in a week or so as it really relied on the colors and lighting, while only having some implied details. That was the starting point anyway. I was also playing Destiny 2 at the time and liked the little environments where you could interact with characters with a nicely framed environment background. The idea was to create something like that, space where the player would be using the computer and looking outwards into some sort of vista.
As I’m sure most artists can relate to, after starting the project, it quickly expanded in scope and size as I started just playing around with different shapes and ideas inside 3DS Max. I didn’t feel the need to re-create the concept 100% but wanted to use the key elements that grabbed my attention such as the vast sense of scale, the color palette, geometric shapes and general composition. Throughout the entire production, I just kept reminding myself that it was the lighting and materials that would be doing most of the heavy lifting. The goal was to just quickly pump out the models and use the scene as a lighting study.
I would estimate total production hours to be somewhere around 60-80 hours, so about 2 weeks of “full-time” work. Momentum is a huge thing for me to actually finish personal work I find. Harnessing that initial excitement of a new scene and getting as many big steps done as quickly as possible really helps. Keeping a smaller scene size or overall scope can be beneficial as well.
There really isn’t anything too complicated going on in terms of modeling. I enjoy making materials, world building and lighting scenes a lot more than spending time pushing and pulling vertices. I also didn’t want to have to deal with any high poly or unique unwraps, as those things tend to multiply production time like crazy. Certain objects like the computer console would probably have benefited from being treated as a hero piece, but I was more focused on the macro (big picture) and not the micro (small details) with this scene. The goal was to just have fun and not have too many “rules” in play.
I started with creating a rough shape of the hallway silhouette and then used that template to build inwards, extruding out the various ribs and cutting in panel shapes. Then I just broke apart the angled surfaces and rotated that template to be vertical/flat to make modeling the various wall panels and objects a lot easier. I also knew I would be using tiling materials and a simple detail/trim sheet for most objects so I modeled with trim placement and UV strips in mind.
Everything is pretty modular and grid-based, I tried to keep it all so it could snap easily and also be re-used/scaled. Most of the scene construction is just mirrored. And then I just experimented inside the editor, moving light fixtures around, and crashing them through geo to find interesting looks. Sometimes just rotating and scaling an object can yield interesting results, especially for far-off vista style objects. I ended up needing quite a lot less unique objects than I thought I would, simply by re-using whatever I could and implying more detail than was actually there through shadows and silhouette.
The computer is actually made up of several different objects, combined into one in the engine, so you could make quite a few variations or tech panels with just a few objects below. With a few extra pieces it could easily be made into an extremely re-usable kit.
I am a big fan of using prefabs so I focused on modeling as few objects as possible, but duplicating them inside the engine and combining them into larger prefab objects such as one side of the hallway which is then mirrored, or one section of the walkway which is then duplicated with all the railings and light fixtures in place. It is a lot faster to work with these larger objects inside the engine instead of having to select hundreds of individual assets over and over should you want to move something (I’m actually currently working on a tutorial all about prefabs).
For the materials, I just made a base metal, chrome and rubber material in Substance Designer that I could tile across most objects. Then in unreal, I just plugged in the Normal and AO of the tiling texture I made for variations of those same 3 materials but with normal and AO detail that could be UV’ed. In 3ds Max, I just used the AO map as a diffuse to help me see where the details were while doing the UVs.
This also leads to the added benefit of having a lot of repeated elements, which is something that tends to make sci-fi or non-organic things like buildings look more grounded in reality. I tried to re-enforce the triangular theme in various assets without going overboard.
I knew I was going to go with a centered composition for my final shots, so having lots of leading lines all flowing towards the AI core helped with that. I tried to re-enforce the triangular shape of the hall with the blue light strip at the end, which also helps frame the AI. I planned the environment to be seen from only a few key screenshots and from a limited players perspective if it was an actual game. Planning out your shots in advance can really help you zero in on where you need to maximize your time and efforts.
I think a big mistake a lot of beginner environment artists make is they focus on trying to make every little detail just right, when most of the time after a lighting pass, a lot of it is hidden in shadow or not even noticeable. This can lead to a lot of mismanaged time and skewed priorities – and this is why blocking out your scenes is so important. Doing a quick lighting pass can give you some context on where to invest your time. 6 hours spent modeling and uniquely unwrapping/texturing a background prop off in the distance is a waste, especially when you can quickly model a few interesting shapes, UV it with trims and tiling materials in 30 mins and have it achieve 99% of the same result.
While the scene feels large in size, it is actually not very complex. For example, I made a “ram chip” that I just duplicated and scaled at random size variations for a lot of the tech details. I find just experimenting with a limited set of assets in Unreal you can get decent results if you use them creatively. As most of the scene was technically a vista, I wasn’t too concerned with strict texel density. Most of the visuals were coming from the lighting, normals, and reflections anyways, not detailed albedos and intricate edge wear in the texture work.
There really is no “animation” really. You will notice that most of it are just things rotating. I knew I wanted to have some motion in the scene to break the static feel, but I am really no technical artist or animator. So I hacked it and just made objects rotate in unreal, such as the fan and AI core (rotates on 3 different axis at once). You can make almost any object spin/rotate in unreal super easily by adding a rotation component and then you just adjust the speed! A cheap and dirty way to get things moving and add a bit of life to a scene. Note: it will only rotate when you play the scene in the editor, not while you are world building/working.
The lighting setup is pretty much based on the concept from beeple. Red in the hallway with a large blue glow at the end. I just placed a blue point light with a fairly large radius at the end, and then used red point lights for the neons in the hall. For the computer console, I added a blue spotlight to rim light it and fake light reaching it from the AI.
Then I just went around and added a few low-intensity fill lights where I felt they were needed. I try to work with larger pools of light instead of having hundreds of smaller point lights near each little light fixture, which can sometimes lead to a chaotic, spotty look.
90% of the lighting is baked in this scene, for the blue fan lights they are just spotlights shoved into the fan socket, with volumetric fog turned on and a high-density value for those particular lights to give the blue light shafts. The overall sci-fi look comes from the use of extreme colors and lots of emissive details scattered throughout the scene. I went for something pretty dramatic and colorful which was a fun change of pace from the art styles I have been working with for the past few projects.
Aside from a tiny post process contrast tweak, and a little vignette, there isn’t much post effects going on. I relied on the color of the lights and not on color grading for this one.
The smoke particles were lifted from Wikors awesome sci-fi scene he released, I just exported them to my scene and scaled them up. He had them set up to be affected by lights so they fit right into the environment with zero work on my side.
I also took the BP-Fogsheet from Epic’s infiltrator demo which is a simple depth fade plane of fog you can duplicate around to add atmosphere and help silhouette objects.
As an environment artist, I have zero issue with using pre-made assets or things like FX made by other artists. I could spend 20 hours trying to learn how to make VFX with varying levels of success, or just leave it up to the experts. Trying to make EVERYTHING yourself from scratch is a great way to quickly lose momentum and never actually finish a project. There is nothing wrong with investing time to learn new skills, but diving into the cascade editor wasn’t one the initial goals of this project for me.
I don’t see any reason why it wouldn’t work as a game environment really. Everything is pretty modular and the scene runs at a decent framerate on my GTX 1060. The only real thing that would need to be done is invest some time in LOD’s and making proper collision for the various objects. As I knew this was a portfolio piece I didn’t bother with those 2 things. A few hours of technical tune-ups and it would be ready to rock.