Geremia Merzari told us about the working process behind the Lost in Nature project, explained how the foliage was created, and talked about the scene composition and lighting.
Hello everyone! My name is Geremia! I’m 26 years old and I’m currently studying Game Art in Belgium. I enjoy spending my spare time working on core skills and learning new workflows aiming to become a successful environment artist in the game industry.
Before starting my journey in Game Art, I was studying for a master’s degree in Translation, Literature, and Culture at UCL, London. During the time I spent in England, I discovered how passionate I am about Literature, in particular the storytelling aspect of it, which I later found out is crucial in Environment Art. While I was writing my dissertation, I started getting myself familiar with the world of 3D. After reviewing several YouTube tutorials and podcasts, I decided to change my career path and dive into this amazing industry. Due to my strong interest in storytelling, I concluded that I should focus on enhancing my skills in Environment and Prop Art.
Besides learning core skills such as modeling and texturing to achieve AAA results, I also decided to strengthen my composition and lighting skills since they are pivotal aspects of Environment Art. Hence, I decided to build this scene to practice lighting and composition while including vegetation and foliage studies, which I really enjoy.
The Lost in Nature Project
The theme behind Lost in Nature is 'Ruins'. I’ve always wanted to create a scene focusing on vegetation and improve my visual storytelling skills throughout the process.
The first step was gathering references. I think this phase is great for brainstorming. While keeping your theme and goal in focus, you can browse through Pinterest, Google Images, and Google Maps to explore and find new sources for inspiration. I tend to gather references from both real life and concept art to explore different moods and ideas. At the same time, it’s important to stay loyal to the real world while attempting to recreate a realistic scene.
The fundamental tool in this stage is PureRef, which is a great free program to organize all your references and to brainstorm. Another add-on I use is ImageAssistant, which helps to extract images from websites.
My PureRef file was centered around Nordic forests, from the trees to the ground, to understand how foliage builds up in real life to then try to reproduce it in 3D and achieve a realistic look.
The purpose of this scene is to enhance my composition and lighting skills while always keeping in mind that the storytelling element is very important.
Specifically, the work behind the lighting on this piece was focused mostly on achieving a certain mood: a utopic daytime forest, where Nature and architecture combine.
The first thing I did to establish proportions in my scene was to build a quick blockout using basic shapes in Unreal Engine 5, only after the focus point was established. From there, I built everything around my point of interest and created a camera to keep in mind what my final renders will look like. Once the blockout was done, I also did a rough blockout for the lights to have a quick idea of how the scene will be lit.
I use SpeedTree for all my vegetation meshes, I barely scratched the surface with this software, but I really enjoy the workflow and the speed that comes with it. It gives you the power to tweak every single aspect of the mesh you’re creating. I mainly use it procedurally, but it also offers the chance to model and sculpt assets.
I start with a trunk for any kind of foliage mesh. I use the Absolute Steps mode in the generation menu to create flowers because it has more control over the stems. Then, I tweak parameters to match the model to the references. The major tweaks I did are in the Generation, Spine, and Segments menus. However, when it comes to leaves for trees, I spend time in the Deformation menu to add extra realism.
Then I added little branch meshes and started tweaking again while keeping the references in mind.
I use the Proportional mode because I can easily control the position of the branches in the generation menu. I experiment with “First” and “Last” in order to move the branches along the stems.
The next step is to apply a leaf mesh to the plant and adjust and tweak the main settings in the Generation menu. Finally, I apply extra random features and increase the readability of the mesh in the Deformation menu.
To add materials to the leaves, I use the built-in material editor and add high-quality atlases from Megascans or from Textures.
Once the model is done, I always like to implement forces to add a further touch of realism to the model, apply a wind modifier, tweak the intensity to my liking, and finally export all the information into Unreal Engine. By doing so, I save time setting up wind forces in UE, considering how quick the process is inside SpeedTree.
One advantage of using SpeedTree is that when exporting everything in Unreal Engine, it automatically generates the materials, which come with vertex color information. That allows to paint in extra details.
To summarize, these are the menus where I spend most of my time.
In the Generation menu, you can decide how many stems/trunks/grass blades you want for the model and adjust rotation, the distance between the different flower plants, or the distance between the flower’s stems. In the Spine and Segments menus, you can apply the mean tweaks. Furthermore, for the leaf meshes, I spend extra time in the Deformation menu, to rotate, fold, scale, curl, and twist them.
To demonstrate the power of SpeedTree, I made a couple of quick videos to showcase how fast one can achieve a decent blockout to iterate on.
Once I’ve exported everything from SpeedTree and gathered extra pieces of foliage and trees from Megascans, I started to build up the vegetation in the scene. To blockout the biggest part, I used the UE foliage tool. That allowed me to place different meshes at the same time while adjusting several settings to achieve the desired result. Mainly, I tweaked the size of the meshes, giving each one a different scale range, to get a naturally random variation. Moreover, another important thing that I did was to give them random spacing to increase the realism. Furthermore, I subtly tweaked the albedo of the meshes to enhance variation while keeping in mind the references.
It’s important to have the PureRef page open on top of your engine or on another screen to pay attention to how vegetation grows and builds up in real-life environments. For instance, the foliage growing in or next to the water stream, compared to the one growing under a tree or deep into the forest, must be built differently. Ergo, that’s where the references help you achieve credibility and utter realism.
One useful tool that I used to build up my scene quickly and have a fast visual result of what a specific part of the scene will look like is the Unreal Mesh Paint tool. I created the Megascans 3 blend material, where I could pick three different substances from Quixel Bridge, and used them to vertex paint floors or walls. I often use it because it allows you to easily blend different materials and tweak parameters, like the falloff between each material or how harsh/soft they blend, to make it more believable. Most importantly, it gives you the ability to quickly paint and adjust water puddles.
For example, I needed a little stream of water in the middle as a guiding line toward my focus point, and the material helped me to quickly iterate and draw water puddles. Plus, I was able to add some subtle animation to the water to make it more believable and immersive to watch.
Moreover, a great way to give more detail to the trees is to take advantage of the procedural damage, which can be applied to the leaves of the Megascans trees. This feature drastically helped to achieve the mood I wanted for the scene, and with the integration of decals, such as dead leaves on the floor, it enhanced the realism of the scene considerably.
This project was focused on improving my composition and lighting skills. Therefore, I did not model the assets in the scene, except for part of the vegetation. So, I opted to take advantage of the Megascans library also considering the fact that the project was set to be finalized by the University in three weeks' time.
Therefore, I started to collect meshes to implement the architectural side of the scene. Luckily, I’ve found a pack with ruins, arches, and pillars that totally fit my scene on the Unreal Marketplace. Hence, I replaced the shape I was using as a focus point with the mesh from Megascans. Once I placed all the meshes I needed, I dove into the material of the main arch and increased the normal strength of the rock material slightly to get a subtle higher definition and details, considering all the different light sources I had pointing at it. However, I suggest not exaggerating with the normal strength.
Furthermore, to achieve a more realistic look and add interest to the scene, I took advantage of decals. I have mainly used moss decals on rocks, logs, twigs, and leaves to add extra realism under the trees. The goal is ultimately to work in a logical way. For instance, by looking at the scene, the viewer can perceive wind and that the season is set in Autumn. Hence, it’s logical to add some dead leaves at the bottom of the trees and sporadically on the main path. The most important thing I’ve learned so far is that when building a scene or a prop, one must always pay attention to placing props or extra details in a logical and believable way.
Once I have achieved a decent result on the first iteration of the scene, I usually start tweaking the composition and experiment with different types of lighting. I often use props as guiding lines toward my focus point and attempt at finding a balance between an artistic and realistic scene. On top of that, I insert different little props that work as guiding lines but also enrich the story of the environment.
Moreover, to blend assets inside the scene, I simply added multiple smaller assets around the bigger ones to hide any possible intersection or any kind of hard edges, which would otherwise give away a fake 3D feel to a realistic scene. I did attempt to use pixel offset and add dithering to the nanite meshes, but at the moment of writing this article, nanite meshes show up black if you attempt at doing so.
Composition and Lighting
Once the blockout is done, the first thing that I do is create a CineCamera Actor to set up the point of view of the final render to have a clear idea of what my final scene will look like from the start. After that, I can adjust the FOV based on the reference, by doing so, I won’t have to waste time taking care of small areas that won’t be visible in the final render.
One resource that I suggest is a showcase video from Feng Zhu, where he goes through environmental composition and particularly talks about the camera position. He gives general guidelines, shortly followed by a deep dive into the subject.
After, I used the cinematic viewport in the editor to help me understand what the optimal camera position for the final render will be and to have a greater sense of space in my scene. From there, I started working and tweaking the composition.
I wanted to avoid very symmetrical approaches, which can be boring for the viewer, I’d rather experiment with asymmetry. For instance, when I had a decent starting point composition-wise, I made a quick render and brought it into Photoshop. From there, I manually divide the scene into 3x3 and examine the position of my focus point and how I could make the whole composition more interesting. Plus, I use the burn tool to place highlights placeholders (which will be later fixed in UE) in order to give attention to all the little pockets of interest in the scene. Once that is done, I paste the photoshopped render in PureRef and adjust the lights in Unreal Engine accordingly.
Another important thing that I did was to apply a black-and-white filter to the render to check where the image values and if the brightest sport corresponds to where I want the viewer to focus.
In this case, the lighting was working perfectly because without any postprocessing I was able to create a subtle vignette effect just by placing props and adjusting the light.
The ultimate goal of this check is to examine the scene in detail and assure that the level of contrast is the desired one. By looking at the black-and-white render, one can confirm that the light contrast between the focus point and the surrounding environment is helping to guide the eyes to the arch.
Furthermore, I took advantage of the vegetation in my scene to create both strong and subtle guiding lines towards the arch. For example, I decided to place two big trees right in the foreground, aiming to frame the scene. This was a huge shift in the scene, and the result helped to guide the eyes towards the arch and not get it lost in the forest. Moreover, one subtle tweak I did was to have the trees tilted in all slightly different angles to avoid repetition as much as possible.
One last pass I did was to bring the render in Photoshop again and thoroughly investigate the composition. I studied where my guiding lines were pointing and noticed that the composition created little areas of interest around the arch converging all the attention there. Thanks to the guiding lines, it’s almost like the eyes are just walking through the path alongside the water stream, confined by the trees and heading straight to the arch.
Lighting-wise, I usually tweak the temperature before I touch the actual color of the light, mainly because that’s how light works in real life. I keep in mind how real-life landscapes look and implement subtle highlights by using spotlights, bounce lights, and rect lights. For instance, the water stream creates subtle bounce lights that hits the bottom of the trees. Most importantly, I use only one directional light, which represents the sunlight in the scene.
Another example is how the directional light creates a gradient on the trees on the right side, which intensifies the warm and daytime vibes.
I’ve always wanted to tackle a scene in Unreal Engine 5, especially since I’ve watched the Lumen tutorial by Polygon Academy on YouTube. Therefore, I took this chance to create my first completely real-time environment, where I was able to experiment with Lumen.
Furthermore, it was a great chance to scrutinize how nanite meshes interact with such lighting technology. The experience was surprisingly smooth, considering how heavy in polygons the meshes were, but this is just another demonstration of how nanite meshes are well optimized for Unreal Engine 5.
I think the most challenging part of this scene was the lighting. I had to go through several iterations to find the right setup, to achieve the desired mood in my scene. Most importantly, I had to strictly follow my references to build a realistic lighting setup with accurate reflections and bounce lights. One fundamental thing was to take advantage of global illumination to not overfill the scene with lights
After the general lighting looked optimal, I did another detail lighting pass, where I added extra details to all different areas of the scene. Plus, I added subtle highlights and bounce lights to increase the realism and the accuracy of the lights’ behavior.
Once the scene was ready to be rendered, I did a test render, imported it into Photoshop, and adjusted the colors to create a unique LUT for my environment. After that, I applied the LUT and proceeded to render the final shots.
I have learned a lot about how Lumen behaves in Unreal Engine 5 and how I can work with it to achieve the result I want. Coming from Unreal Engine 4, I was expecting to struggle at the beginning with the new interface, but it was straightforward, and I was able to work comfortably in no time.
Furthermore, I had the chance to learn a lot about composition and how to build a scene with a specific point of interest. On top of that, I have learned the importance of creating secondary areas of interest in the scene and blending them together to get a realistic and coherent scene.
As always, the most difficult part of the process is lighting. It’s hard to reach the desired mood and display the wanted vibes through lighting. A lot of iteration and technical tweaking is required, luckily, since the scene was built with 100% real-time lighting, the feedback is instant and motivating.
I would suggest to all the artists out there to experiment with Unreal Engine 5, try the new lighting technology, and if someone happens to feel overwhelmed, there are plenty of tutorials and breakdowns on YouTube, but also vast official documentation online, which can guide you step by step to achieve your result.
I’d like to thank the people at 80 Level for giving me the chance to write an article about my environment! If you’d like to know more about it, check out my ArtStation and feel free to send me a message.