Creating a Fantasy Character With a Whimsical Lantern & Intricate Clothing

Introduction

Hi, my name is Yuriy Drobot. I’ve been working with 3D character models for a relatively short time. Before that, I worked in architecture and 3D visualization. At some point, I realized that games and characters were more interesting to me, so I decided to change my specialty. I’m currently working as a Character Artist at an outsourcing studio.

For this project, my goal was to achieve a convincing character appearance. I spent quite a while choosing a concept and eventually settled on a piece by artist Pedro Krüger. I was looking for artwork with some unique feature that would make the character stand out among the many warriors, assassins, adventurers, and so on. What I liked about Pedro’s work was, first of all, that the character is a child or a teenager. Secondly, she has this whimsical lantern, which also serves as a container for a spirit. Thirdly, her outfit looks like something out of a Hayao Miyazaki cartoon: puffy pantaloons, a strange weapon, and an unusual hat.

This concept felt like it had room for interpretation, since it wasn’t highly detailed but rather showed a general direction and some key elements. To be fair, I deviated quite a bit from the original concept, and I’m sure the final character turned out differently than what the concept artist originally intended. But I hope he doesn’t mind that too much.

Pipeline

Nothing is out of the ordinary here. Cloth elements with large folds are created in Marvelous Designer and then refined in ZBrush. The rest of the equipment starts with a blockout in 3ds Max, followed by high-poly detailing (bevels and damage) in ZBrush. Retopology is done in Maya, and UV unwrapping in 3ds Max. I bake the maps in Marmoset Toolbag, then move on to texturing in Substance 3D Painter. Final renders are done in Unreal Engine 5.

First stage: Marvelous Designer. I’ll use the boots to walk through some of the next steps.

The second stage is detailing and refining in ZBrush. The third stage is retopology in Maya, followed by baking in Marmoset. The fourth stage is texturing in Adobe Substance 3D Painter.

First, I start with a basic leather material. Then, I add dirt to the recessed areas and slightly brighten the raised parts. After that, I refine everything manually and adjust the overall color.

On the Artistic Side

I structured my workflow like this: after the blockout, I started working on the face. Then I moved on to the gear elements and kept coming back to the face at regular intervals. When you stare at the same model for too long, your eye gets tired and you stop noticing mistakes. But if you switch things up — face, shirt, pants, back to the face, boots, lantern, then the face again — it allows you to look at it with fresh eyes and spot issues you previously missed.

Also, I just enjoy switching between different tasks. If I get tired of sculpting one thing, I can move on to texturing something else. It’s great that working on a single character model involves all these different aspects at once.

Gear

The lantern-cage

An interesting challenge was designing the lantern-cage. In the concept art, the artist indicated that the top part of the device opens to release the spirit inside, and also showed how this mechanism works. As with the rest of the gear, the concept outlined the general characteristics, but that wasn’t enough for a 3D model. So I had to "engineer" the lantern, borrowing elements from various other devices.

At first, I looked at industrial objects and various Art Deco-style decorations, but later, I realized that the strongest design cues for the lantern came from elements of 19th-century rifles and pistols, especially in the engravings and moving mechanisms.

The main shape, which had an Art Deco-like style, gave the lantern its distinctive silhouette, so I was able to focus directly on the details.

Just for fun, I explored some ideas for how the spirit might move inside the lantern. Here is the final look of the lantern: 

Gun

Next came the handheld mortar in the character’s hand. The original image featured a breechloading mechanism, which looked great and had that retro-futuristic vibe reminiscent of the film Castle in the Sky. I spent a long time trying to reconcile the rest of the weapon’s design with this loading method. I even found references to exotic experimental systems with cartridges and the like.

But in the end, I couldn’t come up with a convincing design that included the central wheel and the break action of the gun. So I went with the next best thing — a wheellock mechanism. It wasn’t as fantastical or unique as the original sketch, but it allowed for a more believable construction (and, importantly, made it possible to keep the wheel on the side of the gun). That way, I could base it on real handheld mortars that actually existed and were used.

I imagined what Pedro’s weapon might look like in its folded state, slightly adjusted the proportions to be less stylized, and came up with this sketch:

Done some adjusting on volumes in blocking first, then moved on to details and wear marks in ZBrush. 

The usual sequence: draft → blockout → high poly → low poly → bake. The gun followed the same workflow, as did all the gear elements, really. Later, when I shared some drafts in the community, someone pointed out that the trigger looked ridiculous and not functional. It was hard to argue with that, so I replaced it in the finished model, even though it meant rebaking all the maps and retexturing the weapon from scratch.

Brooch

Another interesting part of the design was the brooch. In the concept, it’s represented in a fairly simplified way, but it’s clear that it’s not a standard shape — it’s made up of two elongated, paired forms. It’s also placed close to the character’s face, which means the viewer’s eye is naturally drawn to it.

After a long search on Pinterest, I found an antique brooch that could do the trick. I really liked that, in addition to its intricate shape, the pin also had a kind of pattern on it, which immediately gave it more character.

I adjusted the engraving slightly to make it more readable from afar:

Small stuff

As a finishing touch and to add a bit of life, I added a few small details to the belt and glove. Some of them got lost in the final renders, but even when they’re barely visible on screen, I think they help enrich and add depth to the overall look.

To stay consistent with the wheellock gun, I even made a winding key for the mechanism. It’s barely visible in the final renders, but I find it reassuring to know my character actually has a way to prepare the weapon for the next shot. Searching for belt buckles and all the little accessories online was genuinely fun. For me, that’s the most enjoyable and relaxing part of the whole process.

Clothes

Among the more unusual clothing pieces, the puffy pants and boots stand out the most. The pants, in fact, were one of the key elements, along with the lantern, that made me fall in love with this sketch and pick it out from the crowd. From my amateur perspective, the pants and the lantern form the core of this character.

The design of the pants and boots reminds me of a whimsical fantasy cartoon. As expected, I couldn’t find any direct references, so I built the pants using two or three different sources. It was quite a challenge to get the right volume in Marvelous Designer. In the end, I had to create several layers of fabric to achieve the properly puffy silhouette.

As for the boots, the reference already included an intricate seam at the knee, which gave them a distinct personality, along with a cutout at the top (which, I think, is meant to be at the back).

I felt this detail worked well with the rest of the character’s design, so I pretty much followed Pedro’s reference directly here, except for the lower part, which I borrowed from military boots to add more detail.

Colors

Next, I finished the rest of the clothing and rendered the character in an A-pose — no colors yet, but with an ID map. Before moving on to texturing, I wanted to get a general sense of the color scheme to understand how everything would work together as a whole.

In the original sketch, the colors are shown in a very rough way — they mostly just help separate different gear elements by brightness — so I had a lot of creative freedom here.

At first, I used some reference images from Pinterest and applied their palettes to my model, but then I started scrapping one option after another. Choosing between the final three was tough, so I asked my wife and a few friends to vote. Naturally, everyone chose something different, so I decided to put the color decision on hold for a while.

I came back to the colors later, once the materials were ready and all that was left was to set the tones. I created a few more color schemes, and — with some help from friends and acquaintances — picked the one that worked best.

Before texturing, I wanted to see the full character as a whole, so I put together a few proper shots with lighting and posing. At first, I tried doing them in ZBrush, but once I added more than two light sources, the lighting started behaving unpredictably. In the end, I switched to KeyShot, which handled the high-poly models with ease and let me quickly apply procedural curvature and dirt maps. These maps really brought the renders to life.

Those are ZBrush shots. I used the 3D Scan Store sample female head to add the pores. ZBrush renders really made the small details pop, but it was quite hard to light the entire model well as a whole.

Left: ZBrush, Right: Keyshot

I created a few draft poses and shared them with friends and people in the community. When you show your work to someone else, you start noticing new things yourself — things you’d missed before because your eyes had gotten too used to the model.

Hat

That’s how I realized, for example, that the hat looked odd from every angle except the side.

Up to that point, I had been trying to stick to the hat’s original shape from the sketch, but when a few friends pointed it out as well, I decided to redesign it into something more traditional and recognizable.

I completely redesigned the hat. The fun little upward-pointing tip was gone, but the overall shape became much more readable. It was interesting to age the hat in a way that went beyond just adding wear and tear through textures. As an experiment, I tore it in several places on the high-poly model. Then I added stitching threads to hold the seams together and baked those onto separate transparent planes. Some ragged bits and torn patches were added later during the low-poly and texturing stages. I wanted to break away from perfectly clean model edges — and I think, to a large extent, I managed to do that.

Green: a plane, that has only stitches with opacity and is floating above the hat’s surface. A floater. Red: planks with torn fabric and opacity.

If you take a close look, you can clearly see the tricks, but it works quite well from a distance.

A similar approach was applied to the cloak:

Pose

When it came time to pose the character, I ran into the issue that the poses kept looking unnatural. I used DAZ to pose an avatar and then adjusted the clothing to fit. Even the ones that seemed decent at first glance ended up looking awkward when I looked at them the next day. Some subtle, hard-to-pin-down details were giving away the artificiality of the character.

An example of an unconvincing pose:

That’s when I asked my wife to pose for me. To make it easier to position the hands, I quickly put together life-size mockups of the gun and lantern.

After posing the avatar using a photo of a real person, the result looked much more natural and visually pleasing:

Portrait

The most difficult and time-consuming part of the project was definitely the portrait. As I mentioned earlier, I would work on a piece of gear, then go back to the portrait. When my eyes and focus got tired, I’d set the portrait aside and switch back to gear again. Then I’d return to the portrait once more. It’s not the most efficient workflow, and probably not suitable for game production, but for a personal project, I had the freedom to experiment and explore, and I took full advantage of that.

At first, I planned to keep the face stylized and simplified. But with each new iteration, I realized I wanted something more realistic. Eventually, I understood that if I wanted the face to feel alive, I’d need to put in the effort to capture some likeness.

This was clearly more work than the character really called for — especially since the face in the concept is mostly covered by a bandana — but I was curious to try it. I’d never attempted portrait-level likeness before, so it felt like a good opportunity.

Then I just spent a couple of months trying to sculpt a portrait of Sophia Lillis. It turned out to be really difficult. I knew portraits were challenging, but I didn’t expect it to be this slow and tough.

My choice for a first likeness was not ideal: a young girl’s face is very smooth, with no wrinkles, and all the forms are soft and subtle. It’s hard to latch onto two or three distinct features that would carry the likeness.

If I had to do my first portrait over again, I’d choose someone with more pronounced facial features — l ike Danny Trejo or Willem Dafoe. Unfortunately, neither of them suited the character, so I kept crying and trying to get a decent likeness of Sophia.

After several iterations, I settled into a workflow: I would tweak areas in Max or ZBrush, then export the model into UE and render it from angles that matched film references as closely as possible. Then I’d compare the frame from the movie with my shot and mark areas that needed changes, doing this across 5 to 6 angles at once.

The hardest part was figuring out how the soft forms, like cheeks, should shift. The tricky thing is, you never know exactly what lens or distance was used for a shot, and facial volumes in ¾ view can vary dramatically depending on lighting, camera angle, and expression. Then I’d go back to the model and repeat the whole process again. It’s a clunky workflow, only justifiable by my original goal of getting a final render in Unreal Engine. Next time, I’ll definitely combine modeling and rendering in one tool, like Blender, to streamline the process.

Just like that — only with 5 or 6 cameras at once. Despite all the extra effort that went into achieving the likeness, I have to say it was worth it. When you’re sculpting a generic face — taking a little from here, a little from there — it’s much harder to catch mistakes. But once you’re aiming for a specific face, those mistakes inevitably become visible. You simply don’t notice them at first, but the result ends up feeling subtly off.

Hair

I created the hair using cards, following a standard game development pipeline. I placed them manually in 3ds Max. For me, bending the hair cards using an FFD Box is much easier than doing it in Maya or ZBrush, and it gives me a lot more control.

The eyelashes and eyebrows, however, were done as meshes. In this case, mesh was the better choice because I was aiming for portrait-level likeness, and that often meant adjusting both the brows and the part of the face underneath them. Meshes were more convenient for that than hair cards. I could’ve replaced them with cards at the final stage, but there wasn’t much point.

Rendering in UE

In Unreal Engine, I used a simple lighting setup with indirect light and several direct light sources — usually 5 to 8 per angle — to highlight the key elements. I adjusted the lighting for each individual shot. 

I mostly relied on multiple low-intensity lights rather than one or two strong ones. This approach created a more complex and interesting light structure and helped build a mood of darkness and mystery.

Background

At one point, I considered creating an active background for the main shot. I made a few sketches, but ultimately decided against it. The character looked detailed enough, even against a blank backdrop, and skipping the background allowed me to save a couple of months of work on the environment.

Conclusion

This project took me about a year to complete (working roughly two hours a day in my free time outside of my main job). I didn’t expect it to take this long.

By far, the most difficult part was achieving a believable likeness — that alone took about a third of the total time. I can’t say I’m completely satisfied with the portrait, but I think it’s good enough for the scope of this project.

I didn’t follow any specific guide or tutorial — just Googled things as they came up. So I can’t recommend any particular course, but there’s more than enough material on YouTube and Google to answer any question.

Overall, I made this project for myself, and it was a pretty fun experience. And in the end, I think it was worth it as a portfolio piece.

Image credits: All reference images belong to their respective copyright holders.

Thanks: Big thanks to the friends and colleagues who helped with advice and feedback!

Yuriy Drobot, 3D Artist

Interview conducted by Amber Rutherford