How to Create Skin with Micro Details for a Realistic 3D Samurai

Introduction

My name is Rassoul Edji, I am a Texturing and Look Development Technical Director with a broad range of expertise, including modeling, animation, environments, lighting, and compositing. I am passionate about exploring cutting-edge technology, optimizing workflows, and developing efficient pipelines — all to create the most photorealistic results as accurately and quickly as possible. I strive to deliver visually stunning imagery and push the boundaries of visual effects.

I studied VFX at University for 3 years, which allowed me to dabble in the entire pipeline. During this time, I did a lot of projects on the side both for myself and for commercial clients. Those 3 years at University allowed me to really develop my skills and become a well-rounded Generalist.

I was at MPC for the past two and a half years as a Lead Texturing and LookDev Technical Director in the Character Lab department, contributing to Transformers Rise of the Beasts, Sonic the Hedgehog 2, Percy Jackson and the Olympians, and Snow White.

About the Samurai Project

I’ve always wanted to work on a project set in historic Japan. I’ve loved that aesthetic for as long as I can remember. The original plan was to do a short film with full environments, animation, fight scenes, and more. But, that didn’t come to fruition due to members of the team not being able to commit the required time to bring the project to life.

I did still want to make a Samurai, so I kept on working on the character despite the larger project being scrapped. The design and textural complexion of a Samurai character wasn’t something I could pass on.

There is a lot of media around historic Japan and Samurai — movies, games, museum pieces, historic re-enactments, etc. I did a lot of research on Samurai and how their armor was made for the project, and during that time, I gathered a lot of references for Samurai armor from museums, films, and research documents. These served both as inspiration and blueprints for my character.

Head and Face Workflow

The entire human body was designed firstly in MetaHuman Creator. This was because it provided the easiest and most interactive way of designing a human character that was also of a very high standard to begin with. It also came with a rig that I wanted to use for blocking animation in the sequence. This rig could then be passed onto the rigger to upgrade and enhance as needed for our needs, saving precious research and development time. Once I had a design that I liked in MetaHuman Creator, I exported the model to Maya.

My friend Riccardo Gheller collaborated with me on this project. He handled the sculpting of the face and the grooming. I sent the model to him to further push the realism of the face, by adding better details and anatomical accuracy. We had a few back-and-forths discussing what worked and what didn’t work as the model progressed until we reached a point that we were both happy with. The sculpting work was done entirely inside of ZBrush.

The eyes from the MetaHuman model were not up to the standard that I wanted, so I modeled new physically accurate eyes based on physical research papers and dimensions of the human eye. The final eye models were made up of a Cornea, a Limbus, a Lens, an Iris, and a Meniscus.

All of the grooms on the character were made inside of Maya’s XGen. The Samurai had over 320,000 individual strands of hair, making up his head, face, and body. Supporting groom was made for the ropes and the clothing. The groom totaled over 13 descriptions and 508 guide curves.

Make of Armor

The entire armor was modeled based on a reference from photoshoots of real Samurai armor. I made the entire armor in Maya with traditional poly modeling techniques. Each piece was hand-modeled to match references. I was very keen on getting the model as authentic as possible in terms of how it functioned. 

Because of this, I modeled each strap and rope to ensure that it was representative of how the real Samurai armor was strapped down with ropes and bands. This provided an exceptional level of geometric detail. The idea was to also have these ropes and straps rigged and simulated when the character moved in the sequence, which would have provided an even higher level of realism to the character's movements.

I made two different primary panel types, this was so that I could use them when switching out the different parts of the model when creating the different variations of the Samurai for the sequence. The idea here was to be able to re-color and hide/unhide parts of the model to create different characters for use in the film. This would’ve let me create a small team of Samurai from the same hero model, all with their own unique looks.

The final character was made up of over 3236 objects and spanned across 113 UDIMs.

Retopologizing

The project didn’t need much retopologizing except for a few pieces which were done in Maya with the Quad Draw Tool. Some of the pieces that were heavier (such as the dragon head on the helmet) were run through ZBrush’s retopo tools to get a lighter model out of it with quads.

All of the UVs were done in Maya. Since the model is comprised mainly of small details, I would first model one piece and then also UV that one piece. Then, when I would duplicate and layout those pieces, dozens and hundreds of times the piece would already be modeled correctly and would even have UVs already completed.

I ensured that the entire character had a consistent texel density. This texel density was calculated based on the closest shot that I had planned for the film, which was an extreme closeup.

Texturing

Texturing for the character was done in Mari. My workflow starts first in Maya, where I assign materials to the objects. Then I run a script I have that combines all of the geo that shares a similar material. This gives me the model with all the objects combined into their respective materials. I then take this model into Mari, which creates selection sets out of the different geo for use as masks during the texturing process.

I baked my Geo Maps in Substance 3D Painter. I have a custom smart material that takes the bakes and breaks them down into different weights. For example, taking the curvature bake and outputting Curvature Fine, Curvature Heavy, Curvature Super Heavy, and Cavity. These were then taken into Mari to be used there. There was a total of 9 geo bakes used in the character. They were: AO, Cavity, Curvature Fine, Curvature, Curvature Heavy, Curvature Super Heavy, Position, Vary, and World Space Normals.

I like to work as procedurally as possible in my projects. Because I knew, when I started the project, that I was also supposed to create textures for a few other Samurai for the film, I created a setup that was non-destructive and allowed for easy changes to the values, textures, breakups, and offsets in all of the different materials. Working like this allowed me to easily change the look of the Samurai’s textures by simply changing a few exposed parameters, instead of diving in and making a lot of manual adjustments.

The final skin textures that fed into the shader were comprised of 7 surfacing channels: Diffuse, Bump, Bump To Roughness, Coat Roughness, DMFP, Roughness, and Displacement. Each of these channels was created using the provided maps from the scans, along with a lot of adjustments made on top of them, to ensure that they were contributing to the look inside of the shader correctly. A lot of supporting masks were created and used to refine the look further in LookDev.

Because the character was designed for a short film that had fight scenes, I knew I had to include variations which are damaged. I included armor damage in the base look, by chipping the edges of the panels with displacement and by introducing scratches where there would be contact with swords and other weaponry. I painted blood on the face, armor, and katana by projecting blood splats extracted from reference to blood and water squirts.

I created a procedural dirt and dust pass and then painted it on top for extra fidelity. The dust provided some nice diffuse on the armor, and the dirt provided the thicker and built-up dirt look in the cavities and some areas where you’d find more dirt such as the feet.

Rendering

The entire project was rendered with RenderMan inside of Katana. RenderMan and Katana are my favorite duo for any kind of project due to their incredible flexibility, reliability, and capability. 

The character relied on the modular shader system from ILM called MaterialX Lama. This system allowed me to create materials by layering lobes either horizontally or vertically. With this shading approach, I was able to create as complex or simple of a material as I needed. This ensured top-of-the-line performance, realism, and physical accuracy.

I spent a lot of time ensuring that the eyes behaved correctly. It was crucial to get the sub-surface scattering and caustics inside the eyes to match real-life references.

The Bump to Roughness technology was used to keep micro details such as skin pores, fabric strands, and rope weaving visible from all distances. The technology works by taking the Bump Map and shifting its output into the roughness channel when the camera is far away. This ensures your high-frequency details remain visible from all distances.

The lighting setup for the Samurai is quite simple. I used a studio HDRI as a base and then supplemented it with three lights, a key light, a fill light, and a rim light.

I made sure to include a set, in my case a studio environment. This enables the light to bounce around the environment and accurately interact with your character. I then made this environment invisible and substituted it for a simple cyclorama in the post.

There was little in the way of comp for this project. I did some slight depth of field in Nuke, as well as some small color corrections and lensing effects. Nothing too major as I wanted to maintain the raw look as much as possible to showcase the 3D work.

Main Challenges

Although I started the project over a year ago now, I was only working on it every now and again when I had the time. In terms of total time spent actually working on the character I’d say between 4-5 months.

The biggest challenge was getting the eyes to look great and behave as they should under all lighting conditions. I spent more time working on the eyes than any other part of the character. Other than the eyes, it was a real challenge rendering the character because of how heavy and complex it was. In total, I spent 404 hours rendering the videos and images I used to showcase the project.

Hopefully, RenderMan XPU will come soon to alleviate these rendering issues going forward.

Advice

My advice for anyone starting as a Character Artist is to do a character that is challenging but also doable. You want to be able to push yourself to create something that both looks great artistically but is also technically challenging. Try to also do a broad range of different characters. Even if you really love human characters, try to do an alien, a robot, a creature, etc. Each of those different character types comes with its own challenges, and each will teach you something different that you can use for your next character going forward.

Needless to say, it is crucial to have and use references for your work. Even if your character is completely made up, try to ground it in reality by referencing similar materials and designs from real life, and apply those to your character. It will both train your eye as well as provide an anchor for your character in reality.

Rassoul Edji, Visual Effects Artist

Interview conducted by Theodore McKenzie