John Carney talked about the production process of his scene which initially started with a few props for practice, shared his layout and texturing workflow and named his favorite software.
Introduction
Hi, my name is John Carney and I'm a prop artist living in Manchester, England. I first got involved with 3D art as a teenager playing around with various modeling packages such as Blender and SketchUp. I would often try to recreate my favorite props from popular games I was playing at the time and quickly took to it. Even then, I recall having a keen interest in joining the games industry, and thus it quickly became my focus in school.
In 2013, I began studying Game Art at De Montfort University where I focused primarily on environment art and always striving to achieve AAA quality and realistic work. I was surrounded by a ton of talented artists who now join me in the industry and are super supportive, well-informed ex-industry lecturers teaching us the latest pipelines being used in the industry. Being on the course and surrounded by talented peers really helped push me and my art to be industry-ready by the time I graduated.
By graduation, I had begun working for Small Impact Games, an indie game studio based in Leicester. Whilst there, I was primarily an environment artist working on a medieval survival game called the Black Death. I also got to work on several other projects from other studies for some indie games and various projects unfortunately still under NDA.
I then went on to become a part-time lecturer at DMU, teaching environment art on the game art course there for just one academic year before finally, I took a job as a prop artist at Cloud Imperium Games, working on the much anticipated “Star Citizen”.
About the Environment
The origin of the project has some odd beginnings. At my current place of work, we have a very specific workflow that rarely permits high to low-poly baking. Because of this, I decided I would practice my hard-surface modeling and baking skills on some lunch breaks to make sure I didn’t get rusty in my abilities. I began with some models of everyday items like a stool, a radiator, and a metal tin with no real objective other than to practice. The project got put on hold for a while, and I picked it up back up during lockdown as something to occupy my free time with the closure of almost everything in the UK.
Whilst looking at reference for a radiator, I stumbled across a bunch of real 1/35 scale model dioramas of bathrooms. I tend to see these sorts of models pop up a lot on Pinterest, especially when searching for "3d Art", and I always find them to be a great source of inspiration. There seems to be a lot of overlap in the goals of creating these dioramas as with environment art. The level of detail in this work was amazing and seemed to showcase a level of storytelling in the scenes that you just don’t see as much in real-world references. (obviously, it’s important to still reference the real world for other aspects of the project such as material definition, scale, etc.)
As well as showcasing great storytelling in the dressing of the scene, the shots of these dioramas reminded me of the grungy disheveled apartments seen in movies like Se7en and taxi driver. I’m a big fan of the neo-noir aesthetics in these films, and I really wanted to replicate that sort of atmosphere in a project. This basically became the overarching goal of my project. What started as modeling practice quickly became an environment project with the goal to create a detailed scene, dense with storytelling and character. Most importantly I wanted to try and capture the creepy neo-noir atmosphere seen in some of the popular movies I mentioned.
Layout
With my goal set and reference images collected, I began on the layout of the scene. Initially, I only had the assets I had made as part of modeling practice to place. Because of this, I was quite open-minded about how the layout could go. From reference images I had collected, I knew immediately that I wanted it to be a claustrophobic and narrow room. Not only did this help add to the atmosphere, but the small and narrow room really allowed me to focus on making it feel cluttered and full of character, a feat that would have been much more difficult to achieve with a larger room. In terms of the placement of the bigger assets, there aren't really many options for the layout of bathrooms, from the reference I had gathered, I noticed they all seem to be consistent and follow a certain formula of layout, e.g. the sink is almost always near to the toilet.
Like a lot of aspects of environment creation, the placement of assets was a very iterative process, the placement of the larger assets was set from early in the project, e.g. the bath and sink, but a lot of the smaller props I constantly shifted around the environment. I think the goal is always to achieve a “balanced” scene, it’s a difficult concept to describe but you know it when you see a scene with too much noise or interest on one side or in one corner. I wish I had some sort of theory to share regarding how to achieve a well-balanced scene, but I think it’s sense-developed over time and one I’m far from mastering.
At a much more micro level, I try to rotate and skew my assets to not produce a methodical looking arrangement, it's far too easy to fall into the trap of just placing assets with no consideration for how they would be used in the real world. I tried to consider the story I was telling in this scene, who the character that lived there was, why he had certain items in the bathroom, etc.
Modeling
With a few assets that I had already created as practice, I began modeling white-box versions of everything I thought I’d need in the scene. By white-boxing you can iron out a lot of issues you tend to run into later, before having committed all the time and effort of taking assets to final art. This stage is also a great time to get a grasp on the scale of the project. I knew just from white boxing and throwing assets around the scene that to get the cluttered and claustrophobic environment I had imagined I was going to have to create a whole lot more props than initially planned. Even at the later stages of the project when modeling was no longer a focus, I would still have ideas for props that I thought would benefit the scene and go back and model it. I think it's important to always remain open to the idea of visiting earlier stages of the pipeline if it can benefit the scene. Sometimes it can take getting to the lighting or even color grading stage for you to notice you need something.
After my first pass of white-boxing, I knew that to achieve that sense of character I wanted in the scene I was going to have to get more creative with the sort of props I was planning on making. I don’t think it was ever going to be enough to just have a bath, a toilet, and a sink, so I started thinking about the type of person who would live here. Taking a lot of inspiration from the films I was referencing for atmosphere, I settled on a creepy stalker character who hoarded his “trophy” photographs and true crime magazines to the point it invaded spaces like his bathroom.
With my white-boxing mostly finished, I went into creating the final art version of these models. Looking back, I arguably went a little overboard by making a high poly and a low poly for all my assets. I think it’s an important skill as an artist to be able to prioritize and know when you can get away with doing less. This is perhaps a skill I need to work on, although I did manage to get loads of high-poly modeling practice in, which was my plan all along!
When creating my high-poly models, I imagine my workflow isn't much different from how most people approach high-poly modeling in 3Ds Max. I tend to “flip-flop” between using the turbo smooth modifier with supporting geometry to achieve my beveled edges or a quad chamfer modifier with a turbo smooth stacked on top. They both have their places, and the way I tend to decide is on the production of the asset in question. For something that is machined in real life (e.g. a bolt), I tend to lean towards the “quad chamfer” workflow as you can quickly get a uniform bevel around the whole model which is something you would imagine you’d get if something was machined in production. For anything else or where I would like to be able to vary the bevel across the model, I lean towards turbo-smoothing and adding supporting edge loops. For my low-poly, I’ll usually remove the modifiers and just applying some chamfers to my base mesh for areas where an edge will be likely seen.
One neat little trick I like to do with my models is to stack noise modifiers on top of my high-poly models, especially for trying to achieve the battered and disheveled look with my models. The surface noise and random variation you achieve with this can be replicated in Painter too using noise generators in the height channel, but I prefer having the detail there in the high poly. Below I’ve included a side by side comparison of two models with and without the noise. As well as some noise I try and use “FFD” modifiers to try and just skew the silhouette of an asset. These slight variations can add a lot of character to a seemingly boring asset. The only assets I produced outside of max were some high-poly cloth simulations for towels and bathmats, these were all made in Marvelous Designer and then my low poly in ZBrush using the decimation tool.
Texturing
With my modeling done, I began texturing in Painter. A lot like the white-boxing stage of modeling, I try to do a white-box stage of texturing, this usually just involves adding base values for my PBR maps and immediately exporting to the engine. Once the import and material creation stages are done, iteration can be done quickly so I like to set up my file paths as soon as possible. At this point, it’s easy to get back to Painter and slowly develop on your base materials to a point you’re happy with.
I also try to avoid relying on the preview of my materials in Painter, too. I think it’s important to focus on how assets look in the software where you intend to finally render them. In my experience, the difference in how materials are rendered between Unreal and Painter and be jarring, it makes for less of a headache to just focus on the outcome. I tend to flick between the individual channels in Painter, i.e. albedo, roughness, etc. and then review the outcome in Unreal Engine and tweak accordingly.
When texturing, I also find it important to check my material values amongst my other assets, obviously, this is something that can be tweaked as you go on in a project but it’s easy to texture an asset in isolation only to find after placing all your assets that they don’t sit well together. For this reason, I like to always check the different view modes in Unreal 4. In here, you can check everything from the base color to roughness and ambient occlusion for the whole scene. By seeing the assets alongside each other with their values, you can begin to level them out to be consistent across material types (with exceptions of course). I think it’s especially important to hit this balance with a scene’s base color. Too often in a dark and muddy scene such as this, a model that has been textured in isolation can often have drastically different levels of saturation or lightness in the albedo. This unintendedly can draw the viewer's eye to a specific asset, this can be used to advantage to highlight a focal piece for example (think of the bright TNT boxes in crash bandicoot) but it’s important as that its deliberate.
Rendering
I tried to set up my camera shots as early as possible in the production, inevitably things change, and you often stumble across some great shots just flying around a scene. The advantage to having your shots set up initially is that you can build a nice scene around the shot, the opposite of this (and this is something I did a lot as a student) would be to work hard to create a nice-looking scene, only to fly around at the last minute and struggle to find a shot the shows off your work.
When looking for shots, I was trying to replicate the cinematic look seen in my reference to popular neo-noir films. I did a lot of reading into the focal length, aspect ratio, and various camera settings often used in the film industry and thanks to Unreal 4’s cinematic camera actor, it's prettying easy to replicate these settings. Whilst playing around with the many camera effects in Unreal 4s cinematic camera actor, I stumbled across the dirt mask feature and began experimenting with an array of different dirt masks to achieve the dirty camera lens look. Given the context of the environment, I felt it was appropriate and I was really pleased with the result. There’s a whole bunch of dirt masks online that you can test it with and you can even create your own in Photoshop.
With the lighting, it was a very iterative process, and a lot like other aspects of environment art, it seemed to only really come together towards the end of the project. I tried to start with very broad strokes, finding what is too bright, too dark, overexposed, etc. and slowly dialing it back till eventually, I’m playing with decimal values for my lighting intensity and indirect lighting bounce intensity, etc.
The same way I blocked out my modeling in scene I initially just blocked out the man light source, this was the center light in the room. From there, I tried to add as little as possible in terms of extra lights as I often think you can muddy a scene by adding too many. After iterating on the layout of the scene, I added another light source from the hallway, this made for a nice highlight on the wallpaper by the door and helped push the bounce light up for the scene. This was handy as with just one light source in the middle with a focused cone of light, there were a lot of areas of the scene in the corners and crevices of the room where the shadows became too dark.
For post-processing, I honestly changed very little, I recall ramping up the bloom a little bit and really pushing the sharpness, by default I think Unreal 4 scenes tend to be too blurry but like a lot of things, that is just my personal preference (also probably a side effect of me not wearing my glasses). For color correction, I prefer to use a LUT in Photoshop. I know some people prefer doing it in Unreal but for me, it's easier to get feedback and advice when you can share a PSD and have people really dig through it. Having used Photoshop for a long time, too, I am much more comfortable with the tools in there than in Unreal.
Challenges
I learned on this project and was surprised it stemmed out of some lunchtime practice. I really got to hone my abilities in every aspect of the production of this environment, from modeling to baking, lighting, and post-processing. The main challenge, however, was simply to know what direction to take the project. I had no concept for this piece, just references from movies with the same atmosphere, reference for the color scheme, practical reference for each individual asset, I even had a reference for the type of grunge buildup. But still, without an overarching concept, I often felt I had no idea how it should look, only a vague idea in my head. This did allow me to shift and alter the plan as the project went on, but I still believe a strong concept initially would have made for a better piece.