Sundown

Goals

Sundown is my final project at Think Tank and is the culmination of everything that I’ve learned while in school. Forest Mead was the mentor who supervised me throughout this work. The goal of this project was to create a completely developed environment.

So, no matter where the camera was, nothing came across as unfinished or empty, like a level in a game. I had to consider the entire space as a composition, rather than the other classic workflow, by setting up a camera for key shots and working from there.

It also takes way longer. By going through this process I learned more about UE5 and leveled up my skills across the board. I hope as I tell you about my project, you find something insightful that’s helpful for your future projects.

Here are some of the technical parameters of Sundown:

– It’s made using Unreal Engine 5, the meshes are mostly nanite (except for glass) and the renders utilize hardware Ray-tracing/ Lumen.

– My texel density was set to 1k per meter.

– I utilized typical pack channel texture maps style of AO, Roughness and Metallic.

– I worked in the power of 2 for sizing and snapping objects.

– I watched all 9 seasons of Naruto Shonen Jump while I was working on it.

Planning & Inspiration

For inspiration, I found a concept by Wanxing Wang titled Abandoned Market. I liked how chaotic the scene was while maintaining a strong path for the eye. I also thought that the concept left enough open for artistic freedom. Once I decided on this concept there were a few preliminary steps before I really got started.

Reference gathering is a crucial first step, and for me, it’s the most fun. I started looking at images of spaces that had experienced natural disasters such as flooding and fires. Urbex videos on Youtube came in handy.

But ultimately I used a lot of my own reference images. I took photos in different shops and convenience stores and realized that because I am about 6ft tall, I could easily use my body as a reference to approximate sizes, placement and distances of things. I continued gathering reference material throughout the time I worked on this project. I also made a list of all assets, calculated how long I thought they would take to make, and made a size chart of all of them with a simple birds-eye drawing of the scene layout.

My asset list calculated that my project would take almost 8 months. This was obviously not doable, but my goal wasn’t to complete the whole list, but rather to help me keep busy and stay on track.

The size chart made it so I could punch in the dims of an object and get a good start with an accurate cube. From there, I made a grey box, followed by starting on my kit. Another helpful activity that kept me on track was taking down notes. I wrote a daily to-do list, a weekly goal list, and all of my project ideas.

I brought my notebook everywhere in case I thought of something and it kept me motivated by ticking off tasks on my lists every day. It’s also helpful if you’re ever feeling stuck or tired of a task, switching gears and picking up something else off the list can make things feel fresh again.

Nanite

After seeing the video of UE5’s new tech I was excited to try out an ultra-high poly sculpt forgoing a normal map. Focusing on edge damage and silhouette. The problem, to be sure, lies in UVing. I’m not a master of Zbrush, so maybe I’m missing something, but I tried a bunch of different workflows and couldn’t escape topology with wacky tris or spiralling edge loops requiring a lot of cleanups.

Maya’s UV editor can’t unfold high-density meshes, and there isn’t a fast way to cut seams when the topology is not good. I got through it though, coming out with a few finished props, using mostly the camera projection tool in the UV editor. Those props still exist in my scene, but it was enough of a headache that I benched the idea of the ultra-high workflow.

I changed focus for a week or so to work on developing my set of procedural, tiling materials and base materials in Designer. Investing in a different UV editor could have been a good move, but I was unwilling at the time. What I ultimately settled on for my modelling workflow resembled a mid-poly style.

I made sure there was no visible faceting, cranking up bevelled edge loops, and used a normal map when needed. Nanite shines with this work flow and it allows you to add enough geometry to have a solid silhouette.

A good modelling tip is to adjust the stretches of the straight silhouette on a model. Nothing in life is pixel straight. Taking the time to slightly push and pull your model’s contours can be a great step to make a scene feel a bit more natural.

Props

When it came to props, I worked from big to small. Once I got to the smaller props I packed as many similarly themed props as I could in a texture sample. Keeping things organized and grouped in folders by size and also object type is helpful for streamlining things.

I believe that solid UVs are just as important as a good model and strong texture. Anytime I tried to cut corners by doing something like an “automatic UV” it ended up taking me longer to eventually get the quality I needed to move forward. Taking a little extra time to consider the directionality of your shells can speed things up once you bring it into your texture software. Most of my small and extra small props ended up getting scaled up between 10-30%.

This was a strategy to get more value out of each prop and ultimately allows for less work. By having larger props, you can have fewer of them which means your viewer is less likely to be distracted by visual repetition and patterns. It’s only noticeably wrong when compared directly with something not scaled up.

Texturing

When it came to texturing, most of my props got heavily desaturated from their true color between 20-60% and across the board I ended up having to opt for more muted tones vs. the typically saturated nature of things in stores.

This approach minimizes work in post. As long as you stay organized with your texturing layer stack it’s fast to fine-tune and adjusts. Paying attention to color tone and saturation was an important lesson I learned from my mentor, Forrest Mead. It took some time to understand, but once I caught on it was fascinating to identify how color grading is implemented in films to drive mood and the viewer’s attention.

Watching an intro to color grading video is a helpful way to understand this even if you’re not planning on getting into the proper software for it. Early in my texturing of products, I was keen on recreating logos and designs using Photoshop.

After completing a few with this method, I realized this process would be a huge time suck. Instead, I bought a bunch of snacks, lottery tickets, gum, etc., peeled the labels off some canned goods at home, and got to scanning! This gave me a pretty good start, and from there I could mostly focus on the roughness, wear, damage and dirt.

One of the materials I made in Designer was the floor tiles. I exposed a few parameters to create fast iterations and adjustments for my final look. I used an opacity map as a guide in Maya to then cut out the missing tiles on a simple, slightly extruded plane. Substance Designer has a steep learning curve and, while working on this project, I became much more comfortable with it.

Scope

When I started this project I had many goals and ideas that ultimately had to get sidelined due to time. For one, a surprising amount of products/objects in a convenience store use clear plastic packaging.

I have seen some great props using clear plastic packaging rendered in engines like Marmoset, but you’ll notice that a lot of scenes opt out of really using transparency.

I think my glass windows and fridge doors detracted from my scene a little and caused me some headaches when rendering, but I felt I needed them there. I wanted to sculpt more rotting foods, create more object variety, like clothing racks and sunglass stands, have a section caved in, flickering lights, dripping liquids animated—lofty goals!

I wasn’t able to complete those, but it didn’t detract from the big picture, which is what matters in the end. My shaders were all pretty straightforward. I had a master material with an option for emissive masks and basic adjustments, A vertex paint material, a transparency master and a decal master.

Mid-way through I started playing around with the footprint of the scene and ended up tightening things up. Having a solid kit that snaps to the power of two makes iterating a breeze.

I also had to move things around spatially to make it work better for the camera. For example, there is way
more space behind the counter compared to any of my references. This is something I’ve heard is very common in the game industry when it comes to level design.

Lighting

Lighting a scene can be tricky. You have to forget about how much work went into all the details and focus on the composition, mood and storytelling. At this point, I decided to pull the trigger and upgrade my 1060 to a 3070 GPU for the purpose of seeing what ray tracing can do.

Although Lumen looks great, ray tracing gives a softer more natural shadow and although it gave me some more headaches I kept it because I felt like the light casting in was driving the scene. Implementing volumetric scattering and light shaft bloom is another balancing act.

It looks great and builds atmosphere but it also tends to flatten things out. So some of that hard work in the roughness variation gets lost. I was weary of having dust particles in my scene; I find them to be used fairly heavily and it seemed a little too obvious so I opted for simple animations of my sunlight to breathe some life into it hoping to create the feeling of time passing.

Rendering

When it comes down to rendering out sequences, you have to forfeit all attachment to the work and effort you put in and focus on the composition and what it’s saying. This was hard for me, I loved all the details and wanted to show them.

Several assets and an entire area weren’t really represented in my final edit. A good lesson is to pause a movie, scene by scene, and note the camera movement, length of shot, and what the composition looks like.

It’s a great exercise in understanding what makes a good cinematic shot. Also dissecting other artists’ scenes, and watching DiNusty portfolio reviews on YouTube are great resources for understanding what makes a good sequence.

It’s also important to note, although it may seem obvious that for a cinematic there is no reason not to move and adjust things for the shot. For most of my shots, I had to hide in-the-way objects and make barriers and false walls for lighting and camera purposes.

Rendering out sequences feels like going to jail. Hit render and pray. Every time I would leave to run errands while it was rendering I would come home to a crash. So babysitting my machine is what got me across the finish line.

End

Finally, having a few fresh eyes on the project while selecting shots was really invaluable. Experienced artists and friends are always good to get all the feedback and there’s no harm in cold DMing someone for advice. Lastly, before you make your final edit always double-check that nothing visible is floating or crashing.