Florian Elie shared the workflow behind the Temple Jar project, explained how the vegetation was made, and gave some advice to beginner artists.
In case you missed it
More from the artist
Hi, my name is Florian Elie, I'm 26 years old and I work at Shiro Games as a Lead Environment Artist on Wartales. I started working in 3D in 2017 at Aslak Studio, then I worked at Tap4Fun Paris and I joined Shiro in 2020.
I had an opportunity to do another interview for 80 Level in 2020, and since then, I was able to continue working on Wartales, learn new 3D software like World Creator or SpeedTree, and improve my skills in Unreal Engine!
My last personal project was Green Green Grass of Tunnel, which I did to learn Unreal Engine. When I started the Jar Temple, my goal was to use what I had already learned about that software as well as the skills I had learned at Shiro to make a new project that mixed many of my favorite influences. So I wanted to create small lore around a series of dioramas. The jar temple is the first of this series, and I really hope to find time to make more and improve them in the coming months.
The Temple Jar Project
I didn't use a specific concept for this project, but I made a mood board of several important references to me, which are mostly Shadow of the Colossus, some Andrew Porter concepts, Mathieu Bablet's Adrasté comic book, and Kena: Bridge of Spirits, for the rendering style and level of stylization wanted. Based on this mood board, I did some quick sketches and started working with them. I also used a lot of photographic references from nature books, ancient temples, urbex, etc. I find it really useful for the fertility of the imagination and understanding how the vegetation around us is made.
It took me a long time to make this project because my goal was to lay the foundations for a series of several dioramas, so I wanted to have as many reusable assets as possible as well as trim sheets and tileable materials. All assets, including vegetation, were made by me. This was the challenge I set for myself. The main software used are:
- 3ds Max and Maya for modeling, mapping, and retopo.
- ZBrush for the sculpt.
- Substance 3D Designer and Painter for the materials and texturing.
- Photoshop for foliage and texturing.
- Marmoset Toolbag for baking.
- SpeedTree for the foliage.
- Unreal Engine 5.
I used various methods for this project. Specific assets such as the jars are made in ZBrush using boolean and sculpt. I did the retopology in Maya and the baking in Marmoset. I used mainly Orb brushes (Michael Vicente) and the trim smooth border with a square alpha to easily flatten the borders.
For their texturing, I made some basic materials with Photoshop and Substance 3D Designer that I imported into Substance 3D Painter and used with some gradients and AO masks, to which I added some highlights. Their texturing is quite simple.
The temple is made of a mix of trim sheets and some props made for the joints and to bring relief. It is fully modeled in 3ds Max, I just exported it with simple material color IDs so that Unreal Engine could recognize different materials directly once imported.
I gave a lot of importance to the texturing and the different elements that would compose the temple. The goal was to make game-ready dioramas, so I decided to optimize everything and make tileable and reusable materials. I was inspired by different patterns from Shadow of the Colossus that I stylized and sculpted in ZBrush. There are 4 kinds of walls/trim sheets, in different sizes and patterns.
I then baked them on a plane and textured them in Substance 3D Painter and Designer. I used this method because it allows me to have perfect control over the shapes and the degree of stylization, which, in my opinion, is more time-consuming and complex to reproduce in Designer (I may not be good enough at it yet), at least for the specific patterns.
I made a second version of these materials with moss using Designer's Height Blend material so that I could paint them later in vertex paint in Unreal and bring variety to the temple and the compo.
I made some modular props to help build the temple.
One of the biggest challenges was the vegetation, it's perfect, it's one of my favorite parts!
I first looked at the opacity maps of the different vegetation types that I needed on Megascans to see how they were made. Megascans is very helpful, even for stylized scenes because it allows seeing in detail how a plant is. I drew the opacity maps in Photoshop and generated the rest of the maps in Substance 3D Designer and Photoshop.
SpeedTree is one of my favorite software, it is extremely powerful since it allows, when a graph is well done, to create a multitude of different versions of the same plant and even to make them grow on other assets thanks to the mesh force. It's extremely useful software for environment artists!
So I started to work on each piece in SpeedTree. The plants and shrubs are relatively simple, a few nodes are enough to get a cool result with very few polygons. When your graph is done, you can generate a lot of different assets just by clicking on the Randomize button.
For climbing and falling vegetation, I used mesh force on the walls with a direction to make it go up or down. I also used this process to make my ivy climb up the tree trunks and have more natural compositions.
I knew that for my composition, I would need a lot of trees, so I decided to optimize them and not exceed 30k tri per tree (except for the giant tree in the temple, which needed more). A really useful node is the frond node, which allows you to have foliage that follows perfectly the curve of your branches while keeping the same level of optimization.
It is very important to keep in mind that a tree has a tree structure (surprise!) so you have to go from the macro to the micro. This is very important to keep a functional and well-done graph. If the branches are not well done, chances are your leaves are not well done, and therefore the procedural randomization will not work very well. It's very important to understand the architecture of a tree to know how it is built to make it logically in SpeedTree.
To bring some variation to the trees, I also made two degrees of bark wear that I painted with vertex paint in Unreal Engine.
So I made different vegetation kits to populate my scene as naturally as possible. For optimization, I made LODs for the vegetation. There are 3 LODs for the grass, and only LOD 0 and LOD 1 cast shadows. When a tree or a plant is made in SpeedTree, we can automatically export the LODs, which saves time and optimization.
A small example of a foliage composition:
Another important point of the vegetation was the grass. My goal was to achieve the Kena: Bridge of Spirit grass quality with the same degree of stylization. So I broke down my blade of grass for a couple of reasons. First, I don't really like it when the grass has the exact same shade as the ground. It's very convenient to get a Ghibli look, but it wasn't really my goal for this project, so I decided to use the RVT only for the base of the grass to make it easier to blend on the ground and give it a lighter gradient color so the silhouette of the grass would stand out.
The final result
The last important part of the project was technical. To bring an "undergrowth" effect to the scene, I decided to follow this tutorial to make a volumetric fog at ground level in Niagara:
An important element to give life to any scene is of course the wind, I followed this very interesting tutorial to have the uniform wind in the scene:
The most time-consuming part of the implementation was the vegetation. I used Unreal's foliage tool to generate vegetation compositions and make them as natural as possible. I used Unreal's mesh paint tool to bring variation to the assets.
Since I didn't have a specific concept for the composition, I just set up the scene with the temple and placed different cameras looking at the shots that were the most interesting. I didn't have a precise framing to achieve, for personal projects, I like to build my scene as I imagine it and then walk around in it to see what the most interesting shots are.
I asked myself many questions about the mood and lighting I wanted, the sunset, the twilight, a warmer atmosphere, etc., and I finally opted for my initial idea, an undergrowth atmosphere. The lighting is very simple, there is just a slightly orange directional light and volumetric fog for the god rays passing through the trees.
Then I played with post-processing for the final touches and set the auto exposure. It is very important to have good harmony of tints and luminances of the albedos in unlit so that the light is uniform and that post-processing does not make aberrations.
To summarize, I would say that the main challenge for this project was regularity. I spent a lot of personal time in the evening after work, an hour here, an hour there. I made a schedule and listed all the main tasks needed to complete this project in advance, and I'm glad I was able to finish it and that it looks like what I had in mind since the beginning!
My advice would be not to get discouraged when you undertake a large personal project, 3D requires a lot of work. The purpose of a personal project is to be free and to have fun, so don’t put pressure on yourself!
Thank you very much, 80 Level!
You may find these articles interesting