Heavenly King Armour Set
Infinity 27 Bootcamp – My Experience in creating the Heavenly King Armour set.
By Andre da Silva
This blog documents my process for creating game-ready models. I’ll share insights into my creative journey — from initial concept to final, optimized asset — highlighting the steps and iterations it takes to fully realize a model for games.
Getting into the Infinity 27 Bootcamp
In May 2025, I was accepted into the Infinity 27 Bootcamp, a program designed to help graduates and aspiring game developers gain valuable industry experience. It’s a four-week course that simulates a real development pipeline in a controlled, collaborative environment.
I joined the Art Strand, where we were introduced to the full production pipeline: from daily stand-ups and progress tracking in Hansoft, to personal development planning (PDP) with weekly check-ins. Each week, we attended workshops covering everything from sculpting and PBR texturing to engine asset integration, mesh skinning and rigging, version control, and working remotely on a game project.
This experience was incredibly valuable — there were so many tools, workflows, and team dynamics I wouldn’t have had access to otherwise. A huge part of this was thanks to our Lead, James, who guided us through the process with industry-level insight and support.
Week 1: Creating the Heavenly King Armour Set
With the introduction out of the way, let’s dive into my four-week process for creating the Heavenly King Armour Set.
The Concept
The core idea was to reinterpret the traditional depiction of the Four Heavenly Kings. Rather than following the usual East Asian representations commonly seen in popular media, I aimed to draw influence from Tibetan culture, ensuring the design would sit comfortably within the world of Samsara — the narrative universe we were working within.
Gathering References
PureRef Reference
I began by researching articles and imagery related to the deities and their associated lore. From there, I expanded my reference board to include a wide range of visual inspiration — not just for accuracy, but to help spark ideas and define a strong visual identity for the set.
I made sure my references aligned with Tibetan culture and aesthetics. Then, I organized everything into clear categories: accessories, armour, clothing, key aesthetics, jewellery, and so on. This helped me better understand what I was drawn to and how those elements could be combined or reimagined.
To refine my reference board further, I went through each image and highlighted what I liked in green, crossing out anything that didn’t fit the vision. This kept the board focused and prevented it from becoming overwhelming — a trap that’s easy to fall into during this early stage.
Thumbnailing & Early Iteration
After organizing my references into categories, the next step was to begin iterating on the idea — starting with thumbnailing. Jono led a workshop where he showed us how to use our gathered references to explore different design ideas and focus on creating strong silhouettes.
We each produced around 25 quick thumbnails, aiming to capture the overall feel and shape language of the design. This phase was all about visual exploration — seeing what worked, what didn’t, and letting ideas flow freely.
Once the thumbnails were complete, we entered a feedback session with the wider team, including tech artists, character artists, and fellow bootcampers. Everyone was encouraged to mark the silhouettes they thought stood out — usually with a green tick or circle. It was a great way to get a broader sense of which ideas resonated the most.
From there, I took the strongest feedback and developed four refined variations of the Heavenly Kings armour set. These versions then went through another round of peer feedback, helping narrow down the direction even further.
Eventually, we settled on the version that felt the most solid — both visually and conceptually. I did a bit more iteration on that final design, which was then reviewed in a one-on-one feedback session with Jono. As a character artist, he helped break down what was working and what needed adjusting. That final review helped shape a clean, focused concept, which was then used in the blockout phase of the model.
Week 2: Blockout Phase & Engine Integration
After several rounds of feedback on the thumbnails and initial design iterations, we finally locked in the concept and moved on to the blockout phase.
The purpose of this stage was to create a rough bounding box — a very basic, low-poly version of the armour set that could be integrated into the engine early on. This allowed us to keep updating the mesh while it was already in place and re-import everything each time a change was made.
At this point, we focused purely on form and silhouette, using a simple base mesh mannequin to build around. The goal wasn’t detail, but to establish space and proportions.
Once the low-poly mesh was built, we skinned and rigged it — binding it to the base skeleton and weight painting it appropriately. After that, we exported the mesh as an FBX, uploaded it via Perforce, and placed it into the depot where all shared assets were stored for engine integration.
With the mesh in engine, we went through the process of hooking everything up — creating a material instance, linking the mesh to the correct base skeleton, and ensuring it would spawn correctly during runtime. This sometimes involved work inside Blueprints to make sure the assigned skeleton and material were properly connected.
Once the bounding box was in and functioning as expected, we were ready to move on to the mid-poly stage — bringing the design closer to its final form.
Mid-Poly Phase: Refining the Form
The mid-poly phase was where the design started to take more solid shape. We made a few iterations to the mesh — particularly around the silhouette — to ensure it read well from all angles. This stage still gave us a lot of flexibility, allowing adjustments to proportions, experimentation with design elements, and planning for sculpting and engine integration.
We started to add more detail, especially in areas like cloth texture and the general structure of the outfit. While the focus remained on maintaining clean forms, we began thinking more intentionally about how the model would function in-game, balancing visual interest with practicality, especially considering animation and cloth simulation.
At this point, we also began incorporating accessories and additional blockout meshes — not fully detailed, but enough to establish their presence and spatial relationship with the main armour. These acted as placeholders as we moved toward the high-poly sculpt.
I also experimented with Marvelous Designer during this phase to explore cloth dynamics and add more realistic folds to the garments. Some ideas — like the one-sleeve concept, an angular skirt, and a cloth belt — carried over into the high-poly version. However, a few elements were adjusted or simplified to avoid potential issues with cloth simulation in the engine.
This stage was a careful balance of creativity and practicality — refining the design while keeping the end goal in mind: a clean, optimized model ready for final sculpting.
High-Poly Phase: Sculpting the Details
The high-poly phase marked the start of the main sculpting process. Since the low-poly base was already solid, I took a more efficient approach — rather than sculpting every tiny detail in ZBrush, I planned to rely heavily on PBR texturing to bring surface details to life later on.
That said, I still made sculpting refinements where they counted. I enhanced the folds in the skirt and trousers to give them a more organic, weighted look and designed a bespoke golden dragon belt buckle as a striking focal point. Housed within the buckle was a large jewel, sculpted in ZBrush to give it a mystical, ancient relic feel.
A key takeaway during this phase — something I learned directly from Jono — was the use of poly groups. Each individual mesh was assigned a poly group in ZBrush, something I hadn’t previously realized could be maintained through export. These poly groups were then used to generate ID maps, which are incredibly useful in Substance Painter. With ID maps, I was able to isolate and texture specific parts of the mesh much more easily and efficiently. It made a massive difference in my organization and workflow during texturing, and it’s definitely a technique I’ll be using going forward.
Once the sculpt was complete and all the main forms locked in, I sought another round of feedback. After implementing the last few tweaks, the model was signed off and ready to move on to the retopology phase.
Week 3: UVs and Bake Preparation
Once all the pieces were retopologized, I moved on to the UV phase. For this project, we targeted a 4K texture resolution, so I decided to split the model into two separate UV maps to maintain texture quality and give better control during texturing.
UV Maps
In Blender, this meant assigning two different materials: one for each UV map. Even though the model was made up of several pieces, I grouped them accordingly and applied the correct material to each group. This setup made the transition to Substance Painter smoother, as I only had to manage two texture sets, making the workflow cleaner and more efficient.
At this point, I had both my low-poly and high-poly meshes ready. To keep everything organized, I followed a consistent naming convention:
-
MeshName_low or _LP for the low-poly
-
MeshName_high or _HP for the high-poly
This was particularly helpful in Substance during baking, as it allowed for accurate matching of high to low using name-based baking, which reduced baking errors and helped avoid artifacts.
There were still a few challenges — some early bakes had issues, but with a bit of back and forth using cage settings and distance sliders, I managed to clean up the bakes and get a much better transfer of the high-poly detail onto the low-poly mesh.
By the end of this phase, I’d kept the final tri count to just around 15K, well below our budget of 25K–30K, without sacrificing the overall integrity or silhouette of the model. The mesh was fully optimized and ready for baking.
Week 4: PBR Texturing
Once the mesh was prepped, I moved on to the PBR texturing phase. I started with the leather material, building it up gradually. The base began with a simple flat color, then I layered in depth and variation using a combination of paint layers, blur filters, pattern masks, and grunge textures to add natural wear and tear.
Since the design was inspired by Tibetan culture, I paid special attention to the jewellery and ornate gold elements — using custom masks and detail brushes to simulate delicate carvings and ancient engraving. This helped give the armour a unique, hand-crafted feel.
The cloth was treated differently, with a more matte, less reflective finish and subtle dirt and wear in strategic areas.
Throughout the texturing process, I kept toggling the mesh in engine — Unreal — to ensure the materials behaved correctly under the game’s lighting and post-processing effects. This back-and-forth was crucial in keeping the look consistent and polished.
Final Mesh
RGB Variation
Final Touches and Integration
The last phase was all about integration and polishing. I imported the textures into Unity, set up the shader parameters, and refined the material properties like metallic and roughness values.
Finally, I rigged the model fully and ran some animation tests to ensure the mesh deformed correctly without clipping or texture stretching. I also created LODs for performance optimization.
After a final review and a few tweaks, the Heavenly King Armour Set was ready for presentation and submission.
Comments
Post a Comment