Colin Wagner did a nice talk about the production of the way you can use modern scanned assets to quickly build some amazing scenes.
Colin Wagner did a nice talk about the production of the way you can use modern scanned assets to quickly build some amazing scenes.
Introduction
My name is Colin Wagner. I’m a 26-year-old artist from Philadelphia, and most recently worked for 4 years on Enderal: The Shards of Order. I’m currently working on its DLC while I search for opportunities to break into the industry. I’ve been in the modding community for a bit over a decade starting in high school continuing all the way through earning my Masters Degree. I’ll never forget the experiences I had in the mod community and all the friends I made along the way.
Growing up as the son of an Architect, I’ve always had an affinity to the form and function of the structures we inhabit, and working to create virtual spaces became a natural extension of my love for both physical structures and virtual worlds. I came to find my favorite aspect of creating virtual worlds was using world building techniques to tell a story for those willing to observe.
Project
I think it all began in the same way most of my scenes begin. In my down time, my mind wanders imagining things to create. I jump from idea to idea until one piques my interest enough to plant a seed. I begin refining this idea, and as I refine it, the urge to create it grows. I imagined a series of environments with one element in common between them. The presence of a strange heavily geometric but almost alien object of some sort, much like the obelisk in 2001: A Space Odyssey. I wanted to create a sense of mystery by showing this same structure in many different places around the world. I’d like to continue those scenes if time permits, this is just the first.
While that’s often how many of my scenes begin, I like to also like to integrate other elements into the process as well; learning something new, or at least doing something I haven’t done before at a bare minimum. My work is not only a creative outlet, but also a learning experience. Before this scene I had worked with one or two Megascans assets per scene, but I had never created a scene of mostly Megascans assets and I felt as though I wasn’t taking advantage of the tools available to me. I also wanted to use UE4’s new volumetric fog in the scene since I’ve wanted to check it out since first hearing about it. From there I decided to blend both and create a speed scene, a scene created from start to finish in a weekend.
Picking up the assets
For the rocks specifically, I grabbed a bunch of rock assets that were visually similar, made from the same or visually similar kind of rock so that I could maintain consistency along the cave walls. The cave walls are mostly just different rock assets clipped into one another. It’s surprisingly effective and ends up looking way better than you expect it to when done right (hiding the clipping as best you can). If I blended two very different styles of rock (for instance, volcanic rock and limestone) it would not have looked like a cohesive cave wall, so consistency was what I looked for more than anything else. As for the placement of the assets, they were all placed by hand, with attention taken to make everything look as seamless as possible. I started with a blockout made from BSPs, and placed the rock assets over the BSPs. I created more ordered areas like the small path from the right portion of the back wall moving towards the center to give areas that looked more “designed” to complement the more random and unruly chaotic areas. That path specifically also has the benefit of pulling the eye from the right of the screen towards the diamond structure. It is one of multiple elements that all point towards that structure, including the volumetric lighting.
Basic BSP Blockout with no lighting.
Adapting the assets to realtime was surprisingly easy. With dense scenes, polycounts can get a bit intense especially for a computer with a GPU that’s a few years old like mine. This problem is exacerbated by the fact that in order to get the most out of its scans, Megascans assets can be a bit dense, especially their lod0 assets. I could have exclusively used their lod1 or lod2 assets, but didn’t want to give up the up-close fidelity. Instead, I took advantage of UE4’s amazing LOD generation. If you haven’t used it yet and are used to making LODs by hand, it’s basically black magic. It’s surprisingly aggressive, yet has an almost imperceptible impact on visuals. Having UE4 handle the LODs gave me the best of both worlds, the looks of higher poly assets, and the performance of lower poly assets without having to make or even import and tweak Megascans own LOD meshes. Just a few clicks per asset and everything was handled for me. Features that reduce the amount of busy work I have to do as an artist are the features I love the most.
Aggressive LODs provide more breathing room to populate your scene with more assets, while allowing your assets to shine up close.
Smaller details
Once I had the cave walls mostly completed, I turned my attention to the smaller details. I thought about what kind of smaller details I wanted in my scene, and decided on 3 major types. Smaller living foliage, mushrooms, and larger dead trees. These three elements gave me a nice splash of color and variety. Like everything in the scene, I’ve tweaked the materials to my needs to help them integrate into the scene more seamlessly (which I will discuss later). When deciding on how to scatter and place these meshes, I used a combination of logical placement (where would these objects be if this scene were real), and compositional placement (where would these assets look best visually). Examples of logical placement include green vegetation closer to the water, a greater number of mushrooms on the dead log/stump, and fallen leaves in the water (there are trees around the opening to the cave that blow in the wind and give the lighting and shadows some motion). The compositional placement is based on your camera angles, lighting, and focal points, and differs with each scene. This is really where you get to use your artistic eye. The placement of minor assets was done in tandem with the lighting, and they both impacted each other. Overall, much like the rock pieces for the cave walls, everything was placed mostly by hand one at a time. I used the single instance mode of the foliage tool for the majority of my foliage placement. If you’re not working on a massive scene, it pays to be very deliberate with placement rather than leaving it up to the foliage brush to place it randomly. The branches in the water are actually recycled from the standing tree asset, always find ways to creatively reuse assets, especially when you have a tight time budget.
For fun, I added subtle animations to the lighting and water in the scene.
Materials
Because the scene was exploratory, I originally had a lot more assets than I used in the final scene. Having so many assets to experiment with and having to create materials for each can be a huge time sink. To counteract this, I used master materials, one for each major style of object. One for leafy foliage, one for mushrooms, one for objects without moss (since moss uses a fuzz map), and one for objects with moss. In each of these master materials, I built parameters that allow me to easily tweak each asset to my needs much faster than if I had made a separate material for each asset. Keeping in mind I was working on a speed scene, I kept my materials very simple and instead focused on dialing in the areas that make the largest difference to a material, such as albedo and roughness. You can get a material to 90% in 10-20% of the time it would take to get it to 98%, and resisting the urge to get lost in the details is important when on a tight time budget. For example, the rock assets I used are designed for an arid climate, but my cave scene is a damp and humid environment, so I darkened the albedo a bit and reduced the roughness to simulate moisture on the rocks. For a great overview of creating master materials for Megascans check out Jonathan Holmes’ livestream. The megascans materials are already excellent, however, so a stronger focus on lighting is going to be the fastest and easiest method to get scan data to shine.
A simple master material with parameters designed to allow me to tweak each asset as needed.
To complement the movement of the shadows from the animated trees in the scene, I added some movement to the water using one of Epic’s particle effects that I tweaked and changed to my needs. The effect creates some turbulence in the water around where the diamond is. It worked to great effect, but the leaves on the water needed to be animated as well. I found a surprisingly simple solution to this issue through adding wind animation to the material of the leaves. Dialing in the values caused the leaves to follow the turbulence of the water better than I expected. Well enough to work at a distance and up close to a degree, but since it isn’t physics based it won’t hold up to scrutiny.
Close-up of the simple material-based animation to get the leaves to match movement with the water.
Light
When working on scenes, I work on the lighting in tandem with the rest of the scene, rather than building the lighting after finalizing the scene. I do this because often lighting and world building can inform each other. For example, I originally had only one source of light in the scene, from the sky (the warmer color light), which left a large portion of the right side of the scene unlit, which led to the tunnel and second (cooler colored) light. The tunnel offered a great way to put some focus on vegetation assets, leading to a greater density of mushrooms, but more importantly the location of the two trees in the path of the light coming from the tunnel. After the larger fallen tree was placed, I wanted to create a small focal point around the stump and break point on the log, leading to a crack in the wall of the cave that let in a small amount of light to bring the eye towards that detail.
Refining the scene, I made liberal use of the global illumination post process function, cranking it much higher than stock to fill in the bounce lighting to be closer to what I believe is realistic. The volumetric fog is a blast to use, as you have complete control over the effect. You can choose which lights affect it, and how powerful their impact is. I set the primary lights to affect the volumetric fog, and dialed in the amount of influence per light. Bounced lighting and fill lights are not affecting the volumetric fog. The primary light from the sky is a directional light that was set to static for better baked bounce lighting, but then set to movable to allow the shadows from the previously mentioned animated trees at the opening of the cave to create movement in the scene’s lighting. The primary cool light from the cavern on the right is a spot light set to static, and I used two spot lights to fill in areas I thought were too dark even with the bounced lighting, mostly focused on the foreground.
Depiction of the impact of each lighting element on the scene. Notice the impact of exaggerating global illumination.
The above image shows the impact of each major lighting element. The first image is just the sunlight (the directional light) with default bounce lighting. With just this light, the scene is minimally lit and it’s very difficult to see anything that isn’t directly in the light. The right side of the scene is void of almost all light, leading to the addition of the floodlight from inside the path to the right. The second image is both primary lights, the sunlight and the synthetic flood light. The volumetric light shafts are helping keep both sides of the scene interesting, but the vast majority of the scene is still unlit and difficult to see. The third image is the impact of exaggerating the bounced lighting (to 80 from a default of 1). This is a good demonstration of the importance of utilizing bounced lighting. Most of the scene is now lit, and much more of the details are visible. However, the foreground is still too dark, leading to the additions in the fourth image, two fill lights to add some light to the foreground. The largest difference is to the leaves floating on the water, which didn’t necessarily read as leaves before they were properly lit.
How do you make two different types of light sources work together?
I very often blend warm and cool light together when lighting my scenes. It offers a nice color contrast to different light sources and gives objects lit by those lights more color variation. I frequently light using the color temperature system rather than using RGB values. It makes more sense to me when lighting to work with kelvin, perhaps because I originally learned traditional film lighting. One benefit of the temperature system is that warm and cool colors don’t clash when illuminating the same surface, the colors blend together to form a more neutral white color. I intentionally used two very different colors and types of light to visually and thematically separate them as much as I could, with one very obviously being sunlight, and the other very obviously being some form of synthetic light (in this case, a floodlight). As for shadows, the way I prevented the lighting from projecting unfortunate shadows was by making both lights light different parts of the scene. The daylight lights the rear and left side of the cave, as well as providing the majority of the bounce lighting.
The floodlight lights the cavern on the right and the surrounding area on the right. Because the light is coming out of the cavern, it’s only projecting out in a relatively narrow cone into the main room, into mostly empty space in the middle. The only object in the direct path of the light is the diamond, but because of the inverse square law, the flood light is very dim by the time it reaches the diamond and makes very little impact on its lighting (and no shadow). Despite this, I didn’t want to segregate warm and cool colors, and wanted to make the two lights look like they were interacting more than they really were. To do this, I used a few techniques to blend them together. The daylight contributes the vast majority of the bounce light, which influences almost every surface in the scene, placing warm light from the daylight in the areas of floodlight influence. I also colored the ambient fog a cool blue color, which also influences most of the scene, including areas the daylight influences. This allows the warm and cool colors to intermingle more, preventing the scene from looking like it was lit in two sections. However, if you look at image 2 in the lighting breakdown, the two lights interact very little in reality.