Fan Zhang shared the workflow behind the Star-Lord project, explained the texturing process, and talked about achieving a character's likeness.
My name is Fan Zhang, I am a Senior Concept Artist and Art Director. My passion for art was ignited during my childhood when I was captivated by various works of fiction such as Transformers, StarCraft, Red Alert and Quake 3, etc. Those fascinations led me to attend art classes, where I received training in drawing and watercolor techniques. During high school, I devoted almost three hours a day to practicing portrait and still life painting, which helped me to establish a strong foundation in art.
In 2004, I was admitted to the China Central Academy of Fine Art in Beijing, where I bought my first tablet, an Intuos 3. This marked the beginning of my journey as a concept designer. I designed various characters after playing games like or how Warhammer 40k, Battlefield, and Call of Duty, and watching movies and reading novels and comics. I learned Maya and how to make animation at that time. I pursued a bachelor’s degree in Digital Media, specializing in Entertainment Design.
In 2009, I was employed by the Fantasy Art magazine as an editor and later as an art director. That job offered me the chance to interview many top creators in the game and movie industries worldwide, including John Howe, Greg Broadmore, Jordu Schell, and many super-talented masters. These experiences were eye-opening and helped me to figure out the direction of my career.
In 2011, I began pursuing a master’s degree at the Academy of Art University in San Francisco, where I received high-quality training from many instructors, most of whom come from the industry. The course covered all aspects of visual development from concept to game engine.
Then, as a freelance Concept Artist, I worked for various clients, including Activision and Microsoft. Later, I worked on a mobile game called Rebel Fire. In 2016, I joined a VR game company where I worked on a game called Stardust. In 2018, I led an art team in an AR glasses company called Nreal, where we created several series of game demos and showcases on our first-generation AR glasses platform, finally winning the Best of CES 2019 award. Since 2020, I have been leading an art team to develop a new mobile game using Unreal Engine 5.
The Star-Lord Project
Star-Lord, performed by Chris Pratt, is my favorite superhero due to his charming personality, sarcastic sense of humor, strategic skills, and charismatic leadership.
As a developer who wants to handle every aspect of game development, I also have experience in hard surface and creature modeling. I do always feel excited about new tools and methods. My current project involves achieving two goals: first, creating a photorealistic character as a showcase version (which we're going to talk about in this interview). Second, make a playable version in Unreal Engine 5 using the MetaHuman standard (I'm about to do this for the next step).
To begin this project, I collected tutorials and watched various videos on YouTube to determine my workflow. One tutorial that was particularly helpful was Kent Trammell's "Human: Realistic Portrait Creation with Blender."
To achieve a head model that closely resembles Star-Lord, I had to put in a lot of effort since I didn't have scan data from Chris Pratt. I had to spend a considerable amount of time perfecting the details to show his personality.
To accomplish this task, I used Blender, MetaHuman, ZBrush, Texturing XYZ, and Photoshop. Initially, I imported a high-poly mesh into MetaHuman to test the outcome. While the result resembled Chris Pratt to some extent, it still needed more accuracy to fit my expectations.
Consequently, in order to keep the UV set of MetaHuman, I exported the mesh from the software and then projected it onto my high-poly model in ZBrush. Then I modified the accuracy and added more details. In this way at a later stage, I can import it to Unreal Engine 5 to leverage all MetaHuman’s resources for playing the character.
For the eyes, I experimented with two different methods. Initially, I used node-based tools to generate eyeballs and all its affiliate structures, which worked well from a certain distance. However, upon closer inspection, I felt that the sclera and iris needed specific details. As a result, I switched to using photo texture and high-poly models for a more realistic look.
Regarding hair, I also tried two different approaches. Initially, I used the particle system, and the result was good, just a little bit too neat to be realistic. Subsequently, I switched to the new hair curve tool, available after Blender 3.3, which made the process more controllable using Geometry Nodes. However, this method required more sculpting work to shape the hair, but the result was more controllable.
The jacket that I choose is from Avengers: Infinity War.
To begin, I imported his full-body model into Marvelous Designer and focused on getting the shape of the jacket just right. I followed the principle of starting from big shapes and structures and working down to smaller ones. Once I was satisfied with the design, I used Marvelous Designer's topology tool to retopologize the jacket. The belt, zipper, and buttons were created using Blender and ZBrush.
One of the challenges I faced was the need for knowledge about real clothing design and sewing. This is an essential part of the process, especially for creating garments with complex structures. Some of the jacket's unique features contained complicated structures that overlapped and were sewn together, which was difficult for Marvelous Designer to simulate. As a result, I had to go back and forth many times to figure out which structures Marvelous Designer could understand and simulate correctly.
Retopology & Unwrapping
The ultimate goal of this project was to create a playable character in the engine using the MetaHuman standard. To ensure that the UV set meets the standard, I imported a body from MetaHuman into ZBrush and projected the UV set onto my high-poly model. In hindsight, I realized that it would have been easier if I had started sculpting with the MetaHuman model.
For retopology, I used Blender and Marvelous Designer.
I utilized the color map provided by MetaHuman as a base and then modified the details and colors on it to an 8K color map to match Chris Pratt's facial features. An 8K specular map projected from Texturing XYZ would be modified to Chris Pratt’s facial features too. For the clothing texture, I used Substance 3D Painter and set up 7 different color ID settings, which allowed me to test and modify colors quickly.
I use Blender's Cycles for rendering my work. Most of the lighting sources are from HDR maps. To ensure that the material and texture settings work well in any lighting situation, I set up several different HDR environments for render testing. In addition, I added one point light and one area light as additional lighting in the final rendering. For post-production, I only did some cleanup work and adjusted the canvas.
To achieve a realistic character render, I utilize different lighting setups and switch between them. This allows me to quickly identify weaknesses and areas for improvement, making it an efficient way to improve my work.
I spent a total of around 300 hours completing this project. This included learning new software and the functions such as Blender (50 hours), Marvelous Designer (40 hours), Substance 3D Painter (20 hours), ZBrush (30 hours), and MetaHuman. The main challenge I faced was learning too many new tools in one project. Even after completing all the tutorials, I still had to spend additional time testing different settings and results to become more familiar with each tool. I made some mistakes during the process, which is the cost for a better result.
Another challenge was achieving the character's likeness, which relied on understanding human anatomy and having a solid art foundation. Although there is a Blender add-on called FaceBuilder that allows modelers to achieve likeness more quickly, I opted to sculpt it myself to challenge my skills.
To aspiring 3D character artists, I recommend focusing on foundational training, especially in anatomy, proportions, and materials. With strong foundational skills, every tool will become easier to use, and you can achieve your goals more efficiently. Don't be afraid to experiment and take risks with your tools. For example, as a Maya user, learning Blender gave me a lot of inspiration and insights to modify my workflow.
Additionally, developing your concept and design skills will help you become a better artist beyond just mastering the tools.