How a Team of 3D Artists Modeled and Textured an Architectural Structure

Introduction

Hi everyone! My name is Simone Mirandola, and I'm currently a Senior 3D Environment Artist at Reply Game Studios in Milan. My passion for 3D began with early CGI movies, but it really took off when polygons started spreading into video games, which I've loved since childhood. Even in those early, flat-shaded polygons, I saw a world that felt alive, extending far beyond the monitor's edges: a space I didn't just want to see but create myself.

In those years, magazines and books were my primary resources, but suddenly, the internet changed everything, with communities sharing images and knowledge at an unprecedented scale. I felt very lucky to have started my career as an Environment Artist at Ubisoft Milan. It was a unique opportunity alongside fantastic people who literally taught me the craft, an experience for which I will be eternally grateful.

During my time there, I've had the privilege of working on iconic titles and franchises such as Rayman, Beyond Good and Evil, Tom Clancy's Splinter Cell, and Rainbow Six, to name a few. Currently, at Reply Game Studios, I have contributed to narrative-driven titles like the Lone Wolf saga, immersive VR experiences such as Theseus, and, most recently, the creation of environments for the hack-and-slash game Soulstice.

Unreleased Project

Regarding the Unreleased Project (developed in Unreal Engine Editor), it has unfortunately been discontinued. I kindly ask you to keep this in mind when viewing the images. That said, I would like to thank the studio for allowing me to share the work the team and I did up to that point.

Working alongside Valentina Zucchi and Simone Urbano, I was responsible for creating architectural remains shattered by mysterious events, along with the surrounding environments. We designed these as self-contained dioramas focusing on their composition, since these spaces were mostly non-interactive.

The Props

For this project, we aimed to introduce a leaner workflow, seeking the best possible compromise between the visual quality we wanted to maintain, our timelines, the small team I was part of, and the strategic reuse of assets (I know, it sounds like the secret ingredients for a magic potion!).

Our concept artists, Steve Scampini and Angelo Ciervo, provided 2D and 3D architectural views, mostly inspired by real, ancient buildings, which guided us on composition and materials. Since some of these concepts incorporated 3D elements, we imported them into Blender to serve as block-out guides.

Meanwhile, we gathered an extensive collection of reference imagery featuring damaged and destroyed buildings, which we studied to bring a sense of realism and weight to our ruined structures.

We broke down the initial concepts into assets, ranging from massive walls to highly reusable decorative details, and favoured a volume-based approach over precise modular meshes, as we did in our latest game, Soulstice.

By assembling all these props into a Packed Level Actor (PLA), we were able to efficiently create multiple scenarios while maintaining a consistent visual structure. Of course, this approach had its limits: to introduce variations between instances, we had to duplicate PLA for independent editing. Despite this, however, we found this to be a good compromise for our needs. All 3D models were converted to Nanite during import to bypass LOD creation and take full advantage of this technology.

Modeling the Assets

The team seamlessly combined various modeling tools and workflows, selecting the most suitable approach for every stage of the process. Communication and feedback were the key, guided by an art direction bible under the supervision of our Lead, Christian Ronchi. To speed up asset production, we established a set of workflow guidelines. For example, almost all props were modelled in mid-poly without unwrapping.

This meant completely skipping the high-to-low-poly baking phase and forgoing geometry-baked normals (with very few exceptions). Instead, we relied exclusively on tileable textures to build materials that were assigned to all objects sharing the same surface.

Starting with the main volumes, we created destroyed versions in Blender or Unreal Engine using Boolean operations, sometimes through the use of stones processed using photogrammetry, and Blender add-ons like OCD and Cracker for rapid damage iteration.

The assets were then processed via decimation or, if needed, Unreal Engine's Remesh tool to generate a uniform triangulated mesh. Finally, we assigned vertex colors corresponding to the shader's RGB channels to drive different materials and damage variations.

For the stone wall, I sculpted a set of unique assets in Blender and arranged them on a plane representing the UV space, starting from the top and side edges. I duplicated and offset these boundary stones to the opposite sides to ensure seamless tiling, then filled the central area, rotating and scaling each instance to add variety.

I baked the result in Adobe Substance 3D Painter to obtain Ambient Occlusion, Normal, and Height maps. The latter was then used to displace a subdivided plane in Unreal Engine, creating a tileable asset that could be easily duplicated as needed. The same principle was applied to a rounded cube with its top and bottom faces removed, allowing me to quickly fill larger structural breaches within the architecture.

The AO was roughly used in Substance 3D Designer to drive the offset of a stone texture, breaking up its regularity, and blended with other maps to improve it. Finally, the Normal map was applied to restore the fine details lost during the optimization of the displacements.

High-res stones were also baked onto decimated versions in Substance 3D Painter for lightweight instancing, or grouped for faster set dressing. The dirt mounds were modeled starting from simple 3D shapes and roughly sculpted to add volume. Using vertex weights, I localized the intensity of the displacement modifier as well as the level of mesh optimization before exporting to Unreal Engine. This ensured great details only where necessary.

Texturing

I am neither a Shader nor a Technical Artist, so I will focus on describing the major features of the custom shader developed by my former colleague, Dario Mambro, to meet our specific needs. Assets feature a multi-layer shader blending up to three materials via vertex colors or ID maps.

Each layer supports Albedo, Normal, and ARMH (AO, Roughness, Metallic, Height) along with other features and can be individually toggled on or off. This Height map controls the influence of one material layer over another, enhancing the blending realism. Additionally, the effect can be tweaked using height offset and contrast values.

Each layer features selectable mapping modes: standard UVs, world-space coordinates, or triplanar mapping. We used the latter most, especially for larger props, to avoid the tedious unwrapping phase where possible and maintain consistent texel density across assets while varying UV coordinates for adjacent objects.

To address the visible tiling patterns that often affect large surfaces, we used vertex painting for local variation and a 'tile-breaking' function for each material layer. This feature utilizes a small noise texture to randomly rotate and offset the UVs, almost eliminating any repetitive look.

On top of these layers, a world-space Z-axis projection handles sand accumulation, with intensity and coverage adjustable via shader parameters.

Conclusion

While it's not a strict rule, I believe that even for fantasy environments, looking at real architecture is the key. It allows us to apply real-world logic, like weight distribution and structural balance, to our designs, giving them a layer of realism and believability. Our main challenge was delivering visuals that met audience expectations by focusing on scenic impact rather than minute details, striking the best possible balance between quality and quantity.

If we had more time, there would have been many things to refine and add, such as decals to create more cracks, or layers of grime and painted decorations to further enrich the walls, just to name a few.

Unfortunately, these images are all that remains of our work, but to me, they represent so much: challenges, enthusiasm, friendships, colleagues, problems, solutions, and hopes. I'm not a guru, and I don't claim to have much to teach anyone, as every experience is unique and the technical aspects of this work change continuously.

However, there are a few small insights that, perhaps, remain timeless. The learning process never truly ends. While this can feel both exciting and frustrating at times, take a breath and rely on your colleagues and the community to improve yourself. After all, our time is limited. Whenever possible, don't forget to balance work and personal life to recharge.

Pablo Neruda once said: 'The child who does not play is not a child, but the man who does not play has lost forever the child who lived in him.' This quote stays with me always. A constant reminder to seek the joy, wonder, and beauty that can be found even in a handful of optimized triangles!

I hope you found this brief project breakdown insightful. There's obviously much more to cover, but I wanted to keep it concise. Huge thanks to 80 Level for this opportunity, to all colleagues at Reply, and thank you for reading!

simone mirandola, Senior 3D Artist

Interview conducted by Gloria Levine