Turning a 2D Concept of Two-Face Into a 3D Sculpture
Emmanuel Okafor shared the workflow behind the Two-Face (Harvey Dent) project, explained how he modeled the human part, and talked about achieving the burned look.
Introduction
My name is Emmanuel Okafor (Luwizart). I consider myself a 3D Generalist, but I specialize in creating 3D characters, which I have been doing for over 13 years. I graduated with a master's in Information Technology (MIT), but even while studying, I knew I would end up as an artist.
I have always loved creative outlets. In my teenage years, I drew comics and later got into 2D animation. In 2008, the Tinker Bell animated movies sparked my curiosity about how 3D was made and led me into the world of 3D art. I discovered Blender in 2012 and fell in love with the feeling of creating something from an empty scene.
I remember staying up until 6 a.m. just working on projects. I didn't receive any formal art education or take courses as a college student. I mainly relied on YouTube tutorials and breakdown reels from studios. They were very insightful and inspired me to see what was possible if I kept going.
Since then, I have worked on several projects as a modeler, including music videos for American artists such as Justin Bieber, Lil Wayne, H.E.R., and others. I have also worked on short films that might not be instantly recognizable but were cool opportunities for me.
Two-Face (Harvey Dent) Project
The Two-Face (Harvey Dent) artwork was inspired by a concept by Stanley "Artgerm" Lau. I liked this concept because it is a bold, in-your-face piece where you see the subject's raw expression of emotion. It offered an opportunity to capture and translate those emotions into 3D. The fact that it is a close-up shot meant I would have plenty of opportunities to have fun with the details.
Most of the work I do is based on 2D concepts. When translating to 3D, it can be easy to miss the soul and intensity of a concept, especially when working off a single image. For this artwork, since it is an asymmetric character, using a base mesh was out of the question.
I started the model with a symmetric sculpt of the human part just to get the head shape. I smoothed out half of it and then sculpted the monster side. This was all done using Dynamesh and eyeballing the forms. Once I was happy with the "base mesh" for the head, I took it into Blender for a technical pass to align the face and its features to the concept, matching it almost 1-to-1.
This usually "breaks" the model, but the goal is to capture the soul and feel of the concept first; I can worry later about how the form works in 3D or from other angles. I did the same for other parts, like the eyes and teeth. My "special sauce" for approaching this is mere brute force, just shaping and adjusting until I get the right feeling.
I rely on the principles of Shape, Form, and Function, but I also expand that to include Material-Reaction and Details/Texture. For the hair grooming, I used Blender. I sculpted a proxy for the hair to capture the rough shape and volume. After the modeling was done, I did the grooming before any shading or texturing, treating the grooming as an extension of the modeling phase.
The process is pretty standard: I created hair guide curves matching the mesh proxy, generated child hairs, and then applied kink effects like clumping, noise, and curliness.
Retopology
I initially wanted to do a proper retopology for the character, but the asymmetric nature of the asset and the fact that it wasn't going to be animated or deformed made me lean towards using ZRemesher. The process for creating decent topology with ZRemesher is similar to workflows used in hard-surface sculpting to get clean meshes.
Using masks, I generated polygroups for different areas of the asset where I usually want closed edge loops and areas where I wanted to hold the edges. Using density paint, I defined areas where I wanted more detail, like around the eyes, nose, and ears. This produced a nice topology that held the details from the high-resolution sculpt and provided clean edge loops for creating UV seams.
For the UV maps, I went with UDIM tiles to capture as much detail as possible when baking the high-resolution sculpt. When creating UVs, you must check for flipped faces, intersecting points, alignment of rotation to the world, and, of course, texel density. All of this ensures good, clean textures that don't look distorted.
Texture
The most challenging part of the detailing and texturing was the monster side of the face, especially the patterns. I started by basically tracing the patterns from the concept. Then, it was time to make it work with the translated 3D design. I was careful not to make it feel too gross or unpleasant to look at.
Once that was achieved, creating the skin details like pores and wrinkles was quite straightforward. I used two custom skin brushes: one to add pores and the other to add wrinkles and breakup lines. For the texturing process, I first baked out the high-resolution maps in either ZBrush or Substance 3D Painter, depending on how long the bake took.
Substance 3D Painter tends to be faster, but baking in ZBrush is sometimes more convenient. For the skin textures painted in Substance 3D Painter, I used both procedural techniques for overall randomness and variety, as well as hand-painted techniques for areas that needed manual direction.
For lighting, during the look development process, I usually work with HDRIs to set up the materials. For the final lighting, I used a 3-point setup (key light, fill light, and rim light), but I still used an HDRI with very low intensity to provide some environment lighting.
After rendering, I added some ambient occlusion during post-production; I like the contrast it introduces. I then did some color correction and added a vignette to wrap it up.
Conclusion
The biggest challenge for this piece was translating the 2D concept to 3D. Getting it to work from the camera view was the easy part, but the tricky part was making it work from other angles, as I planned to do a video presentation. The hardest part was the human face; it looked good from the front, but from the profile or three-quarter view, it looked off and wasn't anatomically correct.
The solution was to compromise so that it would work in 3D. I used a realistic male reference to get the structure right, and then interpreted that to match the expression and feeling of the concept art. My advice for artists starting their journey, or even artists starting a new piece, is not to get carried away by cool techniques or tools.
We are fortunate to have more access to information, resources, and tools than at any point in history. But at the same time, it is easy to get lost in automation and optimization. A successful art piece captures real feelings and emotions. Whatever medium or method you need to achieve that is the "right" technique, even if it takes laying it down brick by brick.