3d artist Daniel Stok talked a little about the way he approached the production of high-quality 3d with some nice complicated vegetation and neat materials.
Hi, my name is Daniel Stok. I’m 22 years old and I’m from the Netherlands. I’m currently in my third year, studying 3D Visual Art at the NHTV in Breda. My focus is on environment art. I’ve had the pleasure of working as a freelancer on a few projects, with my most recent work being foliage for the game ‘Playerunknown’s Battleground’.
After playing The Last of Us about two years ago I wanted to do an Abandoned environment. So, without really planning something or setting a goal, I just started with creating this Abandoned school with the idea that you could walk around the environment with the 3rd person controller in Unreal.
After a short while working on it, I noticed that the whole idea was too overscoped and not planned well enough. So, I transformed the idea from a whole school to just a house. I made a couple moodboards on how I wanted the house to look, and started off with creating a blockout for the scene.
Even with the downscope, I noticed that my idea of a whole house was still too big, as I couldn’t work on this project full time. Instead of creating a fully fleshed house, with a living room, kitchen, bedrooms and such, I set down to just create a small section which you can see in the final images.
I think one of the hardest challenges was to define what I wanted this project to be. It started as this whole overscoped, almost game like, project which was just beyond what I could do and had time for. With the limited amount of time I could spend on this project, I had to ask myself these questions: Do I want to end up showing a lot? Or do I want to achieve a high level of detail? Basically quantity vs quality.
Setting a goal correctly for next time, was definitely something I learned from this project. During production, questions arose that were hard to answer since there were some fundamental parts of my idea missing.
I found it important that the architectural part of this scene felt realistic. So, for each architectural part, I researched how it would be made in real life. For example, with the wall, I looked at the difference between an exterior wall, load bearing wall, and a partition wall. Each version is layered differently for their purpose. This changes, for example, the thickness of the wall, what material you’ll see when the wallpaper/paint starts to peel off and what I would have to create if I would break it up, which was the case for me.
When I started this project, I also just started out using Substance Designer. Throughout the project, I had a hard time keeping consistency between the textures. They often looked good on their own but not in a combined scene. This inconsistency was mostly due to me still learning the tool. It was important that I didn’t just call something done and moved on, I had to sometimes go back to a substance I made at the beginning and update them with things that I’d learned, or even just redo the whole substance. These changes are noticeable when you look through the older images and compare them with the final result.
I used Speedtree for almost all the foliage, except for the grass and the Ivy. Even though Speedtree has some things set up that can grow things like ivy on a wall, I made my Ivy all by hand in Maya.
Since I worked on this project on and off for almost over a year, I learned a lot in between. This is clearly visible with the bushes I made at the start, when I just started learning Speedtree. I was lucky enough to make quite some foliage for ‘Battleground’, from which I learned a lot about various workflows. So, from what I’ve learned, I decided that the bushes needed a good upgrade. I made sure that the leaves have color variation and curvature and I also corrected some of the older colors I had. In addition, I made sure that the cards I used in Speedtree were crossplanes instead of a single plane. I did still use the old texture for time purposes so it does look like an odd branch. Practice makes perfect ?
For all the leaves, I used photo references to create the textures, but with my recent tryouts with photogrammetry, I want to try some new things to get more detailed leaves.
Interiors are often hard to light nicely just by using the directional light, so it’s quite common to add small point lights where needed to just crush the shadows or add color/light to the scene. I turned cast shadows off and kept the intensity very low (50-150) so it wasn’t obvious there were other lights being used than the directional light. In my case, I baked the lighting which gives me proper global illumination (GI), which made sure I have more light bouncing around deeper in the room. To best see the result of GI, I had to bake on production settings. The downside of baking on production settings is that it would take more than two hours to finish, since I like to increase the lightmap resolution a bit for most assets. Therefore, I often baked on ‘preview’ in between, to get a quick feel of how the lighting in the scene would turn out.
Unreal added light portals in 4.11 to tell the light builder where it should focus on light going through openings (doors, windows). This is quite useful with indoor scenes, like mine, in which you solely depend on the light that comes from outside. This was, unfortunately, not enough for my scene. The biggest light difference was done by cranking up the Exposure Bias in the post processing volume, as it’s the same as your eyes adjusting to a darker area. A bias of 1 would make everything 2 times brighter.
For post processing, I heavily relied on a ‘Look Up Table’, with some smaller adjustments you can do in the color grading section. With the LUT I wanted to slightly reduce some of the blue values and increase the red and green values that I had in my scene. I also made sure that there aren’t any black areas by offsetting the shadows slightly, I kept this number as low as possible as I still wanted to keep a nice contrast between light and shadow areas. You never want your scene to be too washed out.
Since the cinematic cameras and sequencer updates in UE4 I’ve started to spend some more time creating a nice simple camera flythrough for my projects. For this, a good aspect ratio was important to support the composition of a shot, so I looked at some common ratios and tried them out in my scene. One aspect ratio that is used a lot in cinematography is a 2.35:1 ratio. To get this working in UE4 you will need to set the right sensor width and height in the cinematic camera. I chose the sensor width of a full frame (DSLR) camera, which is 36mm which you would have to divide by 2.35 to get the right sensor height 36/2.35 = 15.31914893617021. If done correctly, you would see the sensor aspect ratio say 2.35. This is what I personally use as the base for my shots, it may not be 100% correct compared to real world cameras but it does give me the desired result.