Learn How to Create a Beautiful Dreamy Cyborg in 3D

Introduction

Hi, I'm Ngan, a student from Vietnam who is currently taking the Certificate in Digital Production program at Gnomon School of Visual Effects. My interests in this field are character art, hard surface, and concept design. The project I will be showcasing in this article, Finding Rhythm in the Rigid, is my first exploration of these three areas. 

Before learning 3D, I was a self-taught 2D artist. My first experience with 3D was in 2022 when I made an environment blockout in Blender to screenshot and paint over it. At that time, everything I did in 3D was just to build a more solid perspective or anatomical base for my 2D artwork.

However, when I found out about Gnomon school in a YouTube video, it sent me down a rabbit hole of 3D art, and I found myself better suited for this field – one that requires both artistic and technical skillsets. During the 6-month period of making my application portfolio to Gnomon, I started taking 3D more seriously. I was self-learning other programs such as ZBrush, Substance 3D Painter, Marvelous Designer, and Arnold. It was a very tough process being on my own and using whatever resources I could find online, but the challenges only convinced me that this is a path worth pursuing.

I think that my background in traditional art has helped me a lot with all of my 3D projects – from giving me an artistic standard to helping with the composition, anatomy, and personality in my artwork. On the other hand, I welcome the challenges of 3D software and always try to learn as much as possible down to the foundational knowledge because knowing more frees my artistic vision. 

Inspiration & References

The project Finding Rhythm in the Rigid was made for my Hard-surface Modeling 1 and Texturing & Shading 2 classes at Gnomon. Despite spending a lot of time looking for concepts from other artists, this old sketch that I did kept staying in the back of my mind. 

Unfortunately, it has nothing to do with hard surface. On top of that, visualizing a human was not recommended for my texturing class – it was an introduction to Substance 3D Painter course, and skin is too advanced a subject to be covered.

Those constraints only encouraged me to improve on the idea, and my solution was to make her a cyborg. I decided to keep the flowers and design the cyborgs with three layers (skeleton, muscle, and armor), each partially broken to keep the emotion and fragility of the original sketch. Such decisions ended up bringing the artwork a contrast between organic beauty and complex mechanical parts to the artwork, hence the name Finding Rhythm in the Rigid.

I started the ideation process by doing rough 2D sketches in Clip Studio Paint. This allowed me to focus on the emotional aspect and storytelling without worrying about how the components would fit in 3D space. After that, I went back and forth between the drawing software and Blender to test out composition, lighting, pose, etc. This mixed approach utilizes the speed of 2D and the accuracy of 3D, which I find a lot more efficient than only using either of them. 

Eventually, I had a sketch with enough information to start working in 3D. I did not go further because of the time constraint, but I'd love to finish it as a painting one day. 

At the same time, I was gathering references and categorizing everything I found into overall feels, lighting, technical/structural details, anatomy, materials, and design inspirations. 

Modeling

When I did this project, I was just beginning to learn hard-surface modeling concepts, such as support loops and edge flows, from my class. I had to develop my own workflow mixing Maya, ZBrush, and Blender in my free time to keep up with the scope of the project. 

The program and technique I use depend on the shape I have in mind and the way to achieve it more efficiently. A cyborg with multiple layers seemed intimidating at first, but it really pushed me to develop a flexible mindset for modeling. I'm a lot more comfortable switching between programs and processes now instead of feeling hesitant about which way to approach a shape or object.

The skeleton was all done with poly modeling in Maya and Blender. I either started with a cube or used a plane and added thickness when I was done. I referenced a human skeleton but simplified the shapes with sharp edges and flat surfaces to give it a mechanical feel.

Making the joints was the most challenging part when I had to choose the right type of joint for each part and then add my own design to it without affecting functionality. I modeled a single bone for the spine and duplicated it along a curve using Attach to Motion Path and Create Animation Snapshot.

Using modifiers in Blender was a lot faster for me to create the more rigid parts, such as the head's outer shells and fingers. As shown below, by combining the Solidify with two Subdivision Surface modifiers, I can create complex shapes by just extruding a plane. I would import the mesh with the second Subdivision modifier disabled into Maya, delete excess edge loops, leaving enough to hold the shape, and eliminate clipping if there's any.

Sculpting

The outer layer of the neck, arm, and hip was done in ZBrush so that I could incorporate more organic shapes and have more fun iterating on the design. I masked the shape I wanted on a Base Mesh, extracted it, used an edge loop to add thickness, and sculpted with DynaMesh on.

I don't focus much on details or polishing at this stage because achieving precise, sharp surfaces in ZBrush takes much more time than in Maya.

The muscles were created using the Curve Tube brush with tapered ends. After drawing all the curves, I used the Move brush to adjust their volume.

Hair

The hair was done in Maya using XGen. It was my first time using the tool, and CG hair turned out to be a lot more difficult than I expected. I separated the groom into different descriptions: three main descriptions for most of the hair, one more for the short bang since each hair length needs a different noise setting, and a looser and less dense description for the stray hair. 

I used a combination of Noise, Coil, and Clump modifiers to create break-ups in the hair, giving it a softer and more natural look. 

There is a huge difference between having only one Clumping modifier with maps generated from guides and a variety, as shown below. Nevertheless, additional Clumping modifiers slow down XGen considerably, while Coil causes more clipping, so I only added extras if I could manage it. 

XGen is finicky for many reasons: the way it is set up and how some aspects are very technical, like expressions, while others, such as density, width, and length, rely on real hair properties (density, width, length). Therefore, working at a real-world scale, having a proper scene organization, and being observant will significantly speed up the process. For instance, I'd display Debug Level 1 to see the hair count and match that of the average person (around 90,000-150,000 and varies between races), so I don't have to spend time guessing and adjusting the density.

It took me three attempts to get the hang of XGen and use descriptions and modifiers properly. I spent the most time fixing clippings in the groom and ensuring the hair flowed the way I wanted. Even though the flowy style that I chose feels very effortless, XGen hair cannot collide or tangle like real hair, so the extra attention to mimic that was demanding but absolutely necessary.

Retopology & Unwrapping

The parts that I did in ZBrush were decimated and brought into Maya to be retopologized using Quad Draw. I started by drawing very low poly, making the new mesh almost like a blockout to make it easier to polish.

Instead of sculpting the small details, I modeled them into the retopologized mesh because it is more manageable and can produce clean results quickly. For example, to model the forearm, I completed retopology, cut the mesh to get the thin linings, and extruded some areas for details.

The UV unwrapping was also done in Maya. I like using the 3D Cut and Sew tool because it's quicker than manually selecting and cutting edges. I spent a lot of time on the armor pieces and the skeleton. These pieces are supposed to have hard/metallic materials that will make stretching very obvious.

Texturing

Most of the texturing was done in Substance 3D Painter, but I also used procedural materials built in Maya Hypershade for small details like the flower stems and some wires. 

I had a lot of fun hand-painting the texture of the head. Since I was limited to inorganic materials, the cyborg could easily fall into the uncanny valley. This also means that I have to achieve realism without Subsurface Scattering or color zones, blemishes, pores, etc. I took inspiration from ball-jointed dolls, which look human despite their perfectly smooth surfaces, and gave the cyborg a similarly painted look.

Lastly, I added multiple layers of damage, such as dust, moisture, and scratches. Aside from breaking up the roughness, it serves as a storytelling device – the smudge running from the eyes could be her tears, implying sentience. 

Because I decided to go with a subtle color palette for the body, it was crucial to separate each layer and material by roughness. The goal was to be able to see what was going on just by looking at the Roughness channel.

To achieve realistic glass material, I used the Projection tool in Substance 3D Painter with a broken glass texture to paint the cracks. This allows me to be especially precise around the border at the abdomen, in which there would be a very strong light source.

Rendering, Lighting & Post-Production

It took me some trial and error to come up with the right lighting setup. I had a white Key Light on the right of the character, a warm light emanating from her body, and a cool Rim Light to balance it out. Once I figured out the sources, I added low-intensity supporting lights to boost values in certain areas.

I used V-Ray for Maya to render the scene with separate AOVs for each light, Cryptomatte, z-depth, and Back to Beauty preset in case I need to customize specific values such as reflection, specular, etc., in compositing later.

The compositing was done in Nuke. I learned this software with my friends' help and got to see how powerful it is in improving an artwork. I would rebuild the beauty pass using light selects, then add a background, volume rays, dust particles, animate the warm light, and do some slight color grading. 

I made many last-minute changes in Nuke rather than going into Maya/Substance 3D Painter to edit or re-render. That made the process easier and more enjoyable, almost like adding finishing touches to a painting.

Conclusion

The entire project took me three months – I had two months to finish it for my classes, and then I spent another month pushing the artwork even further. That extra month changed everything. Although I no longer had feedback from my instructors, I could work on the project full-time and receive support from my friends. I gathered more references, painted over the areas that needed improving, and switched from Photoshop to Nuke for compositing. 

Many of the challenges that I faced were related not to software proficiency but to fundamental art knowledge. Lighting is one of them. I spent more time than I should have moving the lights around, adding and removing, etc., without having a clear intention for each light. That was slowing me down, so I had to learn the basics like key, fill, rim, and CG light parameters, as well as study more lighting in film/photography.

Aside from that, XGen took me a while to learn because of how finicky the tools can be. The first steps are particularly challenging, as there are many ways to go wrong with setting up the grooming system. I started out by following documentation and different YouTube tutorials to understand the tools and how each artist works with them. I strongly recommend Jesus Fernandez's playlist, which is both very informative and well-structured.

I wish I had known many things earlier, and they could be beneficial for beginning artists to keep in mind. First, be very particular with references, especially when you are doing your own concept. When taking inspiration from another artist, it is easy to get drawn to everything in their art and forget your original intent. I learned to be strict with this process by grouping references into their use and only focusing on each one's purpose. 

Having a to-do list for the project is also recommended. If there have to be major changes, I'd do a paint-over of the render and pick out what I need to work on. The cyborg consists of many small separate pieces. If I look at the unfinished scene and vaguely think that something needs to be added or fixed, I'd never know what to work on next and procrastinate.

Lastly, I encourage new artists, especially those still in school, to experiment and learn more than what's in a course's content. In three months, I learned a lot of new areas on the side, including Nuke, XGen, and building my own hard-surface workflow. The key is to take calculated risks to have something to fall back on in case the new techniques don't work. Furthermore, there is plenty of support in a learning environment. I'm deeply thankful for my instructors, Tran Ma and David Mooy, who were willing to help me with the extra tools that I wanted to learn. Not to mention my friends, who played an equally important role in giving constructive criticism and sharing their expertise with me.

I hope you found this interview helpful, and I'm open to connecting on ArtStation, LinkedIn, and Instagram!

Ngan Hoang, 3D Artist

Interview conducted by Gloria Levine