Turning a 2D Illustration into a Realistic Two-Character 3D Piece

Introduction

My name is Anna Smirnova. I got into 3D quite spontaneously: at some point, I wanted to change my career direction, and through YouTube I discovered 3D art and ZBrush. I started experimenting in ZBrush, and things gradually took off from there.

What I enjoy most is creating characters. As I work on them, I often develop a certain emotional connection with the character: the longer I spend on a project, the more I start thinking about their personality, mood, and the kind of story that might exist behind them.

Throughout my career, I have worked across different visual styles, from stylization to realism. My current focus is on realistic characters, likeness, and hair creation.

I am especially inspired by films, TV series, animation, and game cinematics — projects where the visuals feel connected to a larger story. Only recently have I started to realize that, in my personal work, I am interested not only in creating a strong 3D model, but also in giving the character some of my own context and using visual choices to suggest a story.

From Concept to Story

The project started with a much simpler idea than what it eventually grew into. I was drawn to a fantastic concept by meltifire: a woman in a metal outfit with a stern, intense expression. I wanted to bring that character into 3D and create a full-body piece.

From a technical point of view, I wanted to focus on several things: skin, metallic materials, a detailed costume, and rendering in Unreal Engine. Over the past two years, Unreal has become one of my favorite rendering tools.

From the very beginning, I had chosen a fairly complex character. The design is almost entirely asymmetrical, with a large number of straps, fasteners, and small unique details. At that point, I probably underestimated how much extra work the asymmetry would create. Because a lot of time passed between different stages of the project, new ideas gradually started to change its direction.

At some point, I realized that simply presenting one well-made character was no longer enough for me. I became more interested in revealing more of her personality through storytelling and through her interaction with another character.

I explored several ideas for the second character: a supernatural entity, a captive, a chained character, and an angelic figure. In the end, I settled on a small female demon. With the addition of the second character, the focus of the project shifted. It was no longer just about the model and materials — I also wanted to explore composition, the relationship between the characters, and storytelling.

Another creative challenge was finding the right balance between the original concept and my own story. I wanted to stay true to the strongest parts of the design without limiting the project to a direct 3D interpretation of it. Once you start developing the scene, context, and relationships between the characters yourself, you suddenly have to make many more creative decisions — and that became one of the most challenging and interesting parts of the project for me.

References and Modeling Workflow

My PureRef board usually looks quite chaotic at first. I collect a large number of images, and as the project develops, I gradually clean it up and keep only the references that are still useful for decision-making or continue to inspire me.

I usually work with several types of references: real-life photos, work by other 3D artists, frames from films and cinematics, as well as close-up references for materials and specific details. Real-life photos help me understand how surfaces and materials behave; other artists' work is useful for studying visual solutions and presentation, while films and cinematics often become references for mood, lighting, and composition.

I almost always start modeling in ZBrush. For this project, I used base meshes for both the body and the head, and created most of the elements using ZModeler, standard ZBrush tools, and Maya. Even though a breakdown can make the process look linear, my actual workflow is much more iterative. I continued adjusting the high poly even after moving on to the texturing stage.

One of the main challenges was the large number of small elements and the almost completely asymmetrical design. Because of this, I had to think not only about geometry optimization, but also about how to make the workflow itself more efficient and avoid spending too much time on repeated elements.

For repeated details, such as rivets, I used Micropoly in ZBrush. I also created some reusable elements at the origin, knowing that I could later duplicate them and place them over the blockout meshes.

Whenever possible, I kept elements symmetrical for as long as I could. The legs, helmet, corset, metal claws, and parts of the gloves were initially created symmetrically, and I only broke the symmetry later during the detailing stage. I also try not to rush out of the blockout stage. It is important for me to lock the proportions, silhouette, and main design decisions before moving on to more time-consuming detailing.

I also built the geometry with the future low-poly in mind. Whenever possible, the first subdivision level was already clean enough to use as a low-poly base. Elements such as straps were often created directly as low-poly and then refined in either ZBrush or Maya depending on the task.

Texturing and Material Lookdev

Most of the detail for the skin, clothing, and metal was built directly into the textures rather than handled through the material shader. A more procedural approach could have been more efficient from a production standpoint, but for this project I wanted more direct control over the surfaces. Because of that, I created part of the skin and material detail in ZBrush and then baked it into Normal maps.

In Substance 3D Painter, I usually start with a base material and gradually add curvature-based wear, AO darkening, color variation, gradients, and microdetail. I did not rely heavily on AO in this project because I already knew the final render would use Path Tracing. At the end, I also used Sharpen to improve the readability of smaller details.

Hyperrealism was not the goal. I was more interested in creating rich, detailed surfaces. Because the character design is already very dense, it was especially important to keep a clear visual hierarchy and avoid turning all the detail into noise. I built the textures from large relationships to smaller ones: value → warm center and cooler outer areas → material variation → local wear → microdetail.

In the Base Color, the warm reddish-brown corset creates the main color focus, while I pushed the metal slightly toward a cooler greenish tone to strengthen the temperature contrast and separate the outer silhouette. The leather elements are also differentiated through temperature and value: the corset is warmer and more saturated, the straps are darker and cooler, while the gloves and smaller details are more neutral and draw less attention. I kept the lower part of the character calmer in both color and value.

The metal, on the other hand, became the unifying material across the whole outfit. I intentionally did not split it into many fundamentally different metal types. I usually started with a base material from my library and then added color variation, dirt, and wear, manually reducing or strengthening certain areas to avoid a procedural look.

I wanted the metal to catch the light well and remain one of the most contrast-rich materials in the shot. Its roughness breakup is therefore quite active, while the leather is generally calmer and more uniform, with only a few local high-contrast areas.

It was important to balance how the metal, leather, and other surfaces reacted to light so they would remain clearly readable next to each other. What mattered most to me was how the material worked in the final shot.

Because of that, texturing and lighting developed in parallel. I would check the textures from Substance 3D Painter in Unreal Engine, adjust the Roughness, Color, or level of detail, and then go back to Painter.

The whole project was very iterative. If a form did not work in the final shot, I went back to ZBrush; if the surface did not work, I returned to Substance 3D Painter. The face is a good example of this. During the first iteration, I did not focus too much on likeness. Instead, I tried to get the head and hair into Unreal Engine as quickly as possible and use that as my starting point for further adjustments. For the skin detail, I used Texture XYZ and then added my own color variation and makeup.

At the same time, I continued refining the face shape and creating the hair in XGen. I wanted to preserve the softer, more delicate qualities of the original concept while still keeping a sense of danger in her expression. At first, the face was sharper, but over time I softened and rounded the forms. To push the dark-fantasy feeling a little further, I also gave her cat-like eyes.

For the second character, I took the opposite approach and intentionally kept the textures much simpler. I used a fairly basic beige material for the body and face, while giving a little more attention to the wings through normal detail, cavity, and subtle color variation.

She was important for the story and composition, but she was not the main focal point, so giving both characters the same level of detail would not have been justified.

Composition and Posing

I started exploring the composition fairly early, even before the final textures were finished. I wanted to understand in advance how the characters would interact with each other. The project includes several final shots, and I approached each one individually. In one of the main shots, the main character looks down at the demon.

Her pose, the camera angle, and the direction of her gaze all reinforce her dominant position. There is a sense of coldness and slight contempt in her expression, while still leaving some ambiguity.
At the same time, the second character was meant to support the main character rather than compete with her for attention.

While working on each shot, I constantly analyzed the viewer's eye flow: what they would notice first and where their gaze would move next. The lighting, the face, the costume details, the second character, and the environment all needed to work together and guide the viewer through the image.

Because of that, I did not use the environment purely as decoration. I added elements where they helped connect focal points or support the composition. I used assets from Fab to build the scene and adjusted the environment specifically for each shot. In one of the shots, I also used a simple plane with an image as the background.

Posing played an important role as well. I created a separate pose variation for each shot in ZBrush using Transpose Master. A full rig would have made the process more flexible, but at that stage, this approach was faster and more convenient for me.

Sometimes the changes were very subtle — slightly adjusting a hand, raising a shoulder, or changing the tilt of the body. But, as they say, the devil is in the details: even these small adjustments helped refine the composition and guide the viewer's eye more precisely.

Lighting and Final Presentation

For me, lighting is first and foremost a process of exploration. I keep testing different options until the lighting starts to support the composition, mood, and story. To build the atmosphere, I used fog in Unreal Engine, a large number of local lights, and light blockers.

In No Mercy, I ended up with several different lighting setups. The first was more neutral and was designed to present the characters and materials clearly.

For the second direction, I wanted to push the stylization much further and move closer to a Diablo-like aesthetic: strong reds, deep shadows, and an almost infernal atmosphere. I did not limit myself to a fixed number of lights — if an additional light helped achieve the result I wanted, I used it.

These experiments eventually led to a separate red final shot, where I spent much more time refining both the lighting and the overall atmosphere of the scene.

I organized the lighting changes for different shots through Sequencer in Unreal Engine. After rendering, each shot went through additional post-processing in Photoshop. I mainly enhanced the contrast, adjusted the color, and locally deepened the shadows where the original render felt too flat.

The animation itself was fairly simple and consisted mostly of camera movement set up in Sequencer. The final stills and video were rendered using Path Tracing.

Conclusion

No Mercy is quite an old project for me. I started it in the summer of 2024, but I did not work on it continuously. There were long breaks of several months while I was taking courses, working on other projects, studying anatomy, and growing professionally.

Because of that, the project gradually evolved along with me. What started as a fairly straightforward character piece eventually became a project focused on composition, storytelling, and the relationship between two characters.

One of the main technical challenges was asymmetry. Many of the character's elements were unique, which meant they could not simply be mirrored or reused. The scope grew even further once I decided to introduce a second character. Another creative challenge was finding the right balance between the original concept and my own story.

I wanted to preserve what initially attracted me to the design while making the final piece feel like more than just a direct 3D interpretation of it. No Mercy gave me the opportunity to bring character art, materials, composition, storytelling, Unreal Engine, lighting, and presentation together in a single project, which is why this piece is especially important to me.

I think one of the most important things is to find what you genuinely enjoy doing and what inspires you. If you are taking on a large and complex project, it helps to decide as early as the blockout stage what you actually want to show in the final presentation — what deserves the most attention and what can be intentionally simplified or left out. Otherwise, it is very easy to spend a lot of time on elements that will barely be visible in the final result.

Another important thing for artists at any level is to look at other artists' work, analyze it, and learn from it — but not to use someone else's level as the main measure of your own progress. After finishing a project, I think it helps to ask yourself a few questions: Is this work better than my previous one? Did I learn something new? Did I enjoy the process? If you can answer "yes" to even one of those questions, then the project has already given you something valuable and helped you move forward as an artist.

Anna Smirnova, 3D Character Artist

Interview conducted by Stephanie Almogabar