Recreating Mountain Shrine In 3D Inspired By Journey Through Japan

Introduction

Hi! My name is David Szieber, and I'm an Environment Artist at Nixxes Software in the Netherlands. I started working here a little over 3 years ago, and I have worked on Horizon Zero Dawn Remastered.

I started my journey in Environment Art 5 years ago when I moved to the Netherlands from my home country, Hungary. I am also a self-taught artist, which is very easy nowadays with such amazing resources as 80 Level, The Dinusty Empire, or ArtStation, just to name a few.

In this interview, I will break down the creation of my latest project, Mountain Shrine, and go through the creation process step by step. I will explain the tools I used and the reasoning behind choosing them. The project started 3 years ago during a trip to Japan, but due to a lack of free time, I had put it aside for a long time and only started working on it a few months ago, on and off, after work hours and weekends.

Getting Started

The idea for the project formed during my visit to Japan 3 years ago. I have visited many shrines and temples during my stay, but by far the town of Koyasan stands out as having the most impact on me. Eastern architecture always interested me as it is so different and unique, with intricate detail and precision. I was studying the logic behind these wooden structures, and it was so complex that the easiest way to understand it was to recreate it in 3D.

I started by looking at real-life references. At this point, I really wanted to understand the structural hierarchy of these buildings and how the different elements support each other before I start adding any style to my main building to make sure I keep my blockout realistic.

Once I was happy with my blockout, I began looking at more stylized architecture to make the shrine more unique and to create an interesting silhouette for the building. For this, I found these very interesting Japanese wood model kits. These were amazing to study how I could add some extra stylized detail while keeping the base of my architecture believable.

Modeling

The initial approach to this project was to utilize Nanite and cut corners, not having to worry about creating low-poly meshes and baking, but in the end, I did revert back to the "old" workflows for assets with materials that have directionality, like wood grains, because UVs can be annoying to work with on very dense Nanite assets. Also, my PC is not very powerful, so keeping files manageable helped a lot with load times and iteration.

Transferring UVs from a low-poly did work quite well, but the added hassle didn't seem to be worth it, and the visual difference was marginal, so I decided to only use it where I couldn't use the blockout for the low-poly.

Regarding organic assets like rocks, nanite worked perfectly, as auto UVs and triplanar materials took care of it all.

Most of the building parts were quite straightforward to make, with the exception of the roof, which was more complex. All the wooden parts received a sculpt pass, but the point of this was to just break up the surfaces and to chip edges, as I knew I'd add micro normals in the shader. As this can be quite repetitive, I decided to try the iPad version of ZBrush so whenever I was travelling, I could use that time to work on these.

For the statue, during my trip, I put money into a gacha machine and got a plastic version of one of the Agyo-Ungyo statues. These statues are the defenders of Buddhist temples and represent life and death. I loved these statues and the history behind them, so I wanted to add the one I had to my scene, so I gave scanning a try on my phone. I went with the MagiScan app, but the results were a bit underwhelming as my figurine was really small, so a lot of the detail was kept in the texture for the scan, which I wasn't going to use.

I still decided to keep the mesh result, but once again took it to ZBrush and did a full pass to clean it up and bring back some of the detail that got lost with scanning. I could have spent more time on this, but I knew it would be quite far from my main composition, and also, I'm not a Character Artist, so I would have had to put in quite some effort for a good result.

Foliage & Scattering

I used Speed Tree for the trees because it was very fast to iterate and create decent results. I did not want any specific hero asset for the trees, so I kept them very simple with two variations and a sapling.

For the rest of the foliage, I had a mixed approach. For the grass, I modelled a few individual blades, then created clusters with different looks, and then used these combined for my final mesh to foliage paint in Unreal. This gave me pretty full-looking grass that didn't look flat from certain angles, like cards often do without a bunch of shader tweaking.

For the tall grass, the flower, and the leaves, I did a high-poly version quickly in Maya that I baked down into an atlas. This atlas was then taken to Substance 3D Painter to create the textures, and I made simple assets out of it with cards.

The scattering was done with the foliage paint tool in Unreal, and I also made some clusters of leaves to place around in corners and along the path. I find it very relaxing to put on some music and go through the environment manually to add small bits of detail and story to the scene. It is also very fast to iterate, and the more tasks I can keep inside Unreal, the better.

Materials & Texturing

To start, I wrote up a list of materials I would need and started from the most complicated to the simplest. Most of my materials were done fully inside Substance 3D Designer, except for my wood material and the ground material with leaves. For the wood, I sculpted a tiling texture in ZBrush and took that to 3D Designer, and for the ground material with leaves, I referenced the atlas I created previously for the foliage.

For texturing my meshes, I used the material layering system in Unreal and made RGB masks in Substance 3D Painter. This is a pretty standard workflow at this point, but basically, you use the RGBA channels of a texture to store masks you bake in Substance 3D Painter, and in the shader, you use these to drive where your materials appear on the mesh.

The main benefit of this workflow, in my case, is that I have cohesion throughout the entire project as all assets use the same material library, and it is very quick and easy to make adjustments to my entire scene.

For my master material, I made some basic controls for things I felt would be useful, such as tiling, saturation, normal intensity, etc. I also added a top-down projected moss and set up vertex paint, which, if you are working with Nanite, has to be accessed differently than before. Previously, you had to use the Mesh Paint tool, and now it is under Modeling – Attributes – Paint Vertex Color.

Composition & Lighting

For my composition, I had the same idea from the beginning. A shrine on elevated terrain with stairs leading to it. I wanted the camera to look up at the shrine to make it feel more important and sacred. I divided the main composition into horizontal thirds and used the stairs as leading lines and the lanterns to direct the attention to the temple. I wanted the temple to stand out, so I made sure the rest of the environment didn't have very complex shapes or colors that would draw attention immediately. I also made the temple symmetrical to make my focus point easy to read, but broke this symmetry with small details like the ropes to give it character. I also added depth of field and a vignette to direct the attention even more.

My lighting setup included very little artificial lighting. I only added a couple of Point Lights around the lanterns to separate them from the background and two Rectangle Lights on the ceiling inside the shrine. For my main light sources, I used a Directional Light and bumped the indirect intensity to 4 to bring out all the detail on the underside of the roof elements.

For my Sky Light, I used an HDRI inside Unreal. To access this, you have to make sure you enable the plug-in first. I chose the "Misty_Pines" texture as it had a good amount of blue that worked really well to balance out the warm tones coming from my Directional Light. That's it, basically, for my lighting. I used the volumetric clouds, which worked pretty well and added a hint of saturation in post-processing for my final renders.

Closing Thoughts

For me, this project was an opportunity to practice all the different aspects of environment art and to try new things. It feels really good to have finished it, as I started the project roughly 3 years ago, but it was put aside for years. I count the total time spent on it to be roughly 4-5 months in total on an off after work and on weekends and definitely the main takeaway was to understand how to structure my time as I was only able to work on the project in short 1-2 hour blocks a night so the work I put in had to be meaningful.

If I were to start again, I would spend less time on iterating on smaller details at the early stages, as the project went through some major changes about halfway through. I would also make sure to trust the process more, as it is always difficult to keep yourself motivated when the project enters that stage where the return on hours invested starts to drop significantly.

For anyone reading, I hope you got something out of this breakdown, and feel free to reach out if you have any questions about the topics I touched on.

For now, thank you very much for your attention, and I wish you all the best!

David Szieber, 3D Environment Artist

Interview conducted by Emma Collins