Ian van der Mijn did a breakdown of his Forgotten Shrine made with UE4, Maya, Substance Painter, and SpeedTree for the ArtStation challenge Feudal Japan.
Hi, I’m Ian van der Mijn and I come from the Netherlands. I graduated from IGAD, NHTV in Breda this July and after an internship and summer job at Larian Studios I am now looking for new opportunities. During my study at IGAD, I realized that props and environment art were the directions that I wanted to go. Everything inside me gets exciting when I get to work on a new prop.
At the time when my summer job ended, the ArtStation challenge had already begun and I knew that joining this competition would be a perfect way to get some exposure and hopefully find a new job. So you could say that the main reason I joined the challenge was to show the skills that I acquired over the past couple of months and, hopefully, end up with a very nice new piece for my portfolio.
The first step in setting this project up was deciding which concept I liked most. The main reasons why I chose the concept by Vincent Lebocey was because I had to implement many things that I had almost never done before, and that meant there was more room to learn and improve.
After I picked the concept I started gathering references from the internet. This was going to be the base from which I would take inspiration and build my scene.
I looked at the concept and decided to break it down in larger parts first, starting with setting up the terrain in UE4 and blocking the scene with BSP meshes. I figured this would be a quick and easy way to get a general idea of where the scene had to go and it gave me the opportunity to place a camera and try to match the composition of the concept. BSP meshes are very simple to use and you can easily add and remove them from your scene.
I knew that the focal point of the concept was the figure standing underneath the Torii and I also figured that, since I’m not a character artist, I needed to find another way of guiding the viewer’s eye through the scene. The main idea was to use lighting to allow the first temple to be more of a focal point than the other two. This meant that I had to use different lighting than the concept and that would probably also affect the mood, which was fine by me.
Building the Assets
Building the assets was very exciting. I had experience with creating props since this is what I was doing most of the time during my internship and I felt most comfortable with this part of the project. I started by modeling individual temple pieces in Maya, exported them to ZBrush for the high poly and then back to Maya for low poly and UVs. Once I had all of these pieces done, I combined them and created the modular parts for the temple. With the temples completed, I removed a couple of faces and optimized them a bit better. I knew that the lighting was going to be dynamic so I did not bother creating lightmap UVs.
The roof of the temple consists entirely out of meshes, five different meshes to be exact. These meshes are duplicated, scaled and rotated to get rid of most of the tiling. I created a vertex paint material in UE4 so that I could then paint moss on top of the tiles. This was not the most optimized way of working but considering the time that I had, it was the way I was most familiar with.
In the meantime, I also started texturing the temples. The way I did this was by exporting the smaller, low poly pieces from Maya to Substance Painter and created the wood and roof textures. The plaster walls are a tileable texture that I also created in Substance Painter.
I knew that texturing the temples this way meant that I could not add dirt or damage on the bottom of the temple or wear and tear on the roof so I figured that this would be done with decals inside of UE4.
All of the other props like lanterns and Torri were done using Maya for base mesh, ZBrush for the high poly, Maya again for the low poly and UVs and Substance Painter for the final textures. All of the props, except for the walls that build up the base of the temples.
For the base of the temples, I started out by sculpting a couple of rocks and building them into a wall in Maya. After importing them in UE4, I figured that this did not give me the result that I wanted and switched to a tileable texture with just simple meshes. This way, the walls would be a bit less visible and blend better with the rest of the environment.
This was one of the scariest parts of the project since I had built a tree in SpeedTree once before, three years ago. Also, the foliage was something I had done only two times before. But nevertheless, I wanted the quality to be the same as the rest of the scene. I kept postponing the vegetation because of this and ended up with little time. But this also put me to think about quicker ways of creating the vegetation and decided to take my phone (Samsung Galaxy Note 9) and go outside to find some good reference. Once I decided which types of plants would match the Japanese setting best, I took some pictures (using a blue background is something I discovered, in the end, to work very well) and used these to create leaves and twigs. As you can see, not the best images but I made them work.
For the trees, I started from scratch in SpeedTree. It took a couple of iterations to get the desired result but I’m happy with how they turned out. I sculpted the bark in ZBrush and used pictures again for the needles.
As you can see, this is when things really started to come together. Some of the trees in the background are not as pretty as they could be. The reason for this is that I had no clue how to use LODs for them and they were pretty heavy on the scene. In the end, I used exported billboards for the background. This was a lot easier on the scene but unfortunately not as pretty.
The lighting is a big part of pulling everything together. I used a minimum amount of props and by playing with light and shadow this does not stand out. At almost the end of the project, I completely re-did the lighting and this was a very good decision.
The lighting is fully dynamic, which allowed me to constantly play with the shadows and use this to create interesting compositions. The trees, for example, look a lot more realistic, simply because there are some shadows falling on the bark. I used one directional light, a skylight and a couple of point lights for the candles. I used the light in the candles to create a contrast between cold and warm colors.
To create a better atmosphere I used Unreal’s volumetric fog in combination with a simple fog particle system. All the lights are volumetric as well and influence the fog. The mood is very different than the concept but it allowed me to play with composition and guiding the viewers eyes.
I think one of the biggest flaws in my project is optimisation. I must admit that the FPS count could be a little bit higher but I simply had no time to fix this for the scene, seeing that I only had five weeks to work with. In the end, I only needed to render out a video and take some screenshots so I could get away with this. Knowing that, I decided to spend more time in the overall quality of the assets and lighting.
Like I said, a lot of aspects of building this environment were new to me and this resulted in the whole project being quite a big challenge, but also an even bigger learning experience. I am already looking forward to my next project!