Recreating Sekiro's Shinobi Prosthetic Arm With ZBrush & UE5

Introduction

Hi everyone, my name is Alina M. García Hernández, I'm a 30-year-old Prop Artist with 3 years of experience in the video game industry. When my professional life started as an Interface Designer, the world of video games looked like a distant dream. I wanted to get into video games so badly but I only found app development and marketing jobs. It wasn't until 3 years later that I found myself stuck and then the COVID-19 pandemic hit. I took advantage of that span of time to learn everything possible about 3D and realized that I could be at it for hours without feeling the time passing by.

I was lucky when I was able to step into augmented and virtual reality just as digital marketing companies were experimenting with potential products involving it. I took as many jobs as I could to make up for all the time I felt I had wasted walking in the wrong direction until I finally met people who were in the video game industry and encouraged me to apply.

When I joined the first video game company, I was living the dream while working on various projects. I started with making stylized props that helped me learn the complete workflow, then I began designing scenarios with modular structures, participated in big hard-surface projects for simulators, had the opportunity to create props for Remnant II, and even began to have people in charge and organize projects while training them and even review tests for new entries.

The whole experience led me to empty my portfolio and start over to be able to reach my goal: get into horror video games and be a Prop Hero Artist.

References

When opting to make Sekiro's prosthetic arm, my mind was set to challenge myself to mix the organic with the inorganic, something that would use a wide variety of materials working together for the same purpose as a whole. I love that game and I'm always looking to add meaningful pieces to my portfolio.

First, I researched for references; I have a huge PureRef, filled primarily with gameplay captures and cinematics mixed along with fans' ideas and cosplayers who have recreated the arm. This was to get more of an idea about the details that are not easily observable. Then I look for references of the behavior of each material: how they break, how they get stained, and how they rust or get scratched, adding videos and images about how they're made while taking into account the era of its manufacture. Metal objects nowadays are not the same as the ones smithed by a Japanese blacksmith from the Sengoku era. The difference in details definitely lies in the tools used for its manufacture, among other factors.

To finish the references stage, I always choose a quality bar based on props from other video games that inspire me and are at least more challenging compared to my previous work. For this prop, I chose Hunt: Showdown, I studied the different levels of attrition that were taken into consideration as well as the information about their processes that they've shared with the public, so I could set an achievable goal.

Modeling

The modeling phase started by creating a blockout of all the components in Blender. After that, I was able to identify each piece that will make up the complete model: how they're joined together, and their proportions so they could match the main concept. This was helpful for me because it was easier to determine how much time I was going to allocate to each piece, thus resulting in better time management.

The first doubts to clarify with those blockouts were about how everything was integrated by using ropes, how were those ropes arranged, how many knots, and where they were needed.

I created a Vertex Path and used Solidify and Subdivision modifiers so I could edit it in real-time and thus modify the thickness and trajectory of the entire rope. My goal was for them to look functional, tightened while pulling and thus prevent them from looking overlapped.

Once I had all the pieces roughly defined, I used ZBrush to start working on the high-poly. Perhaps the most time-consuming and detail attention-demanding piece for me was the sculpting of those metal braces fitting around the arm. They may look simple and primitive but deciding how to model them so that the edges didn't have a hard cut, having a bevel where the light could hit and reflect while showing an amount of damages that correctly matched the references was something that I was modifying during several future stages.

That's why I always use layers. I save a Morph Target, add an exaggerated amount of damage, and then clean it with the Morph Brush tool while leveling it with the layers. Even before the final texture, I ended up making some changes.

Modeling the bandages was a whole new experience. I didn't want it to look like a generic fabric. I wanted the wear and damage to be seen, so I used a method that I learned from the artist Pepe Salguero.

I created a mesh that I could subdivide evenly in ZBrush, then I created a Height Map and an Opacity Map in Substance 3D Painter to apply it to the high-poly of my bandages within ZBrush while keeping their UVs well unfolded so that there's no stretching in the model. For me, those bandages served more than anything to correctly fit the prosthetic to the whole body, and on the hands to have more grip when holding objects.

As for the hand, I had many ideas. For a moment there I thought about using bone for the whole hand but then I took into consideration that finger bones with a mechanism inside, without muscles and other organic parts, were tiny hollowed pieces too fragile to withstand combat, so I opted for metal.

I put together the model using Blender with the ropes passing through each piece of the fingers and used ZBrush to add details by combining Milad Kambari's brushes alongside Slash3's and the Damstandar's.

The leather and bones were the things that I enjoyed the most while modeling, even when the bone, no matter how much I researched, I couldn't find which animal or creature it matched. So I experimented with various porosity profiles. The issue was defining how far I should go in the modeling phase to leave the rest to the texturing phase.

I looked for various alphas of bones on Megascans and edited detailed photographs of porosity in Photoshop to create new ones. For the skin, I generated the UVs automatically with the UV Master within ZPlugin and applied a combination of Displace Maps with Surface Noise to then level it into layers, where I divided every type of damage and detail.

Topology & Texturing

I worked the entire retopology in Blender and ZBrush. For most of the inorganic pieces I could go to the lowest Subdivision and I had the piece almost as I started my blockout.

That left me with little cleaning and optimization work, but for the organic pieces many times I tried to get a ZRemesh from ZBrush to work but ended up finding it less troublesome to work on it using Blender from the beginning. I used Sergey Tyapkin's ZenUV to maintain the same densities in the texel or at least it helps me with the overlap process just the necessary to use the UVPackmaster 3 to accommodate them.

For texturing, I settled in by doing the first bakes in Marmoset Toolbag 4. For example, I take out several variants of Ambient Occlusion where I filter what contact data I want for each element, and then in Substance 3D Painter I edit them to create a single final AO map. Once I have my final base maps, I import them into Substance 3D Painter and divide my piece into various folders so I don't get lost with so many textures, especially when I'm about to finish and I feel that a piece isn't integrated correctly.

As for metallic parts, I had a wide variety of colors in references, so I decided that there would be only 3 metals with different tones. I used a combination of Steel Dark Aged with Steel Ruined and I played with the saturation and hue of their base color tones, I modified layers and added small unique details to each one. For example, the piece of metal where the strings are threaded had tiny welding details and scratches with a Grunge Scratches Fine, unlike some of the most exposed metal pieces that withstood deeper hits.

For the bone, I used a cylinder-shaped Warp Projection that I arranged so that the veins could be seen flowing around it. This can also be solved by straightening the UVs from the beginning, but I was able to solve it with the projection and correctly placing the pivots in the image.

At the bottom of the bone's head, I took the same Alphas that I created for the high-poly and used them as brushes to highlight some details in a Height Map and, as a final detail, I added a subtle scattering to the bone so that it could be felt organic.

Dirt & Rust

I divided the dirt and damage variations into two types: ones suffered by each part individually and those suffered by the model in general. I started from the individual type to the general type; each metal has its own way of rusting, or the fabric has stains of absorbed sweat and blood, which is different from the dust that covers the entire piece, the mud that gets into the cavities or the oil stains on the moving parts of the mechanism.

I also subdivided the latter according to its adherence to the material. For the way the fabric absorbs the oil I used a fill layer and another one for the oil's behavior with metallic pieces. Each one started by placing a fill layer of a saturated color that allowed me to identify it with a Black Mask with a Grunge to add variety. I worked on the way it adheres with a mask editor in multiply, I selected whether it goes more in cavities or corners and detailed the areas to my liking by adding a paint or some fill layer in linear dodge or subtract.

I worked the rope as a trim so that all the pieces could have a suitable space on the map. For the rope didn't look repetitive, first I worked the longest piece as a guide and all the others could move around that trim in case I needed to vary any attribute. This prevents the UVs from sharing the Ambient Occlusion map since the contact information would be repeated, so I created a second channel in them to add it within the engine.

As a final detail, I added a system of particles in a blender with three different tris that I textured based on the final colors of the string and took out their Alpha, distributed them with a variation of rotation and size, respecting the pivot so that they were generated around randomly but always with the same end touching the string, and then exported them as a separate element ready to add to the Unreal Engine file.

Adding those details made the prosthetic arm look as a whole integrated and unique piece thus avoiding contrast, such as a brand new polished hand with an old and damaged prosthetic forearm, but I also prevented the dirt and dust layers from looking too uniform on the entire asset.

What I sought was to create variation in these layers by contrasting and defining each part so that it doesn't flatten the roughness thus creating a better composition. You choose which parts to highlight and not get lost between materials.

Rendering

From the beginning of the project, I was planning that the asset would be real-time rendered and would be capable of being integrated into a game engine. I used the PBR Metallic roughness workflow and decided to use Unreal Engine 5, due to its lighting capabilities and easy construction of scenes to render.

The first step was to export all the elements and integrate them into a default scene. There I connected the geometry to a simple shader where I could do a quick test of what the textures looked like and make sure they had the correct formats. At this stage, there was a going back and forth between 3D Designer and Unreal Engine to make different adjustments that allowed me to get the best out of the whole piece, and once I was convinced, I started with the lighting scene.

I did it with the HDRI backdrop plug-in; this is how I started looking for an HDRI that had a good light balance. Usually, this is where I choose the scene's temperature and how much the tones will be affected depending on the lights. I set up a simple plane as my renders' background with a neutral color, and then I added a Post Process Volume, which helped me establish the exposure values ​​and post effects.

This helped me to avoid metallic parts with their reflection being overexposed. I kept their levels between -2 and 2, enabled the scene to be illuminated in real-time by ray tracing in both reflections and global illumination, then I made use of three-point lighting, focusing on highlighting the most interesting parts by placing a main light, another contrast one and a third one behind to illuminate the edge and have a rim light.

I did this for each shot since there were important areas to illuminate and interesting details to show for each arm position. I usually create a cinematic camera and adjust its focus and depth values. Later I place animation keys to the camera and to the final product, in this case, the prosthetic arm, for different positions.

This is how I work on different captures, focusing on being able to get the best out of each one and at the same time being able to return to the previous one in case a change is required. Lastly, I made a simple animation of the hand and placed it in a turnaround to prevent the piece from remaining static.

I think that usually when estimating the time you will invest into a personal project you should always allocate a good amount of it into its presentation. This is the moment when you usually go back to its texture or geometry to improve some aspects that don't benefit your project's final images. This takes time but is worth the effort.

Conclusion

I believe that the better you understand each part of your prop, the further you can take it. It isn't just modeling it and placing it somewhere. We have to look at more details such as the forces they receive, the forces they exert on other pieces, the information that they convey, the silhouette, how the materials evolve over time, all those details that give life to your prop are what we find attractive to the human eye. That's where the importance of references comes in place.

For a beginner, sometimes researching for useful information seems to be limited. My advice to everyone is not to just search for something vague like '3D Helicopter', but look for models through different eras, construction manuals, and avoid trying to invent something because of lack of information. That's what breaks the visual flow the most in a model, even if your piece comes from a fantasy you have to create a background for it. The more passionate you are, the better the story you could create for it.

Alina Garcia Hernández, 3D Prop Artist

Interview conducted by Theodore McKenzie