Recreating a Howl's Moving Castle Character with ZBrush & Marmoset Toolbag

Introduction

My name is Daniel Merticariu, and I've been a 3D Character Artist for eleven years now. I've worked on a lot of projects, with a lot of indie studios, and some triple-A. My preference is towards indies, because I feel they do the most creative work. Although, as it usually happens with most creative endeavors, where it's the most creative, it's also the riskiest. Which, for artists such as myself, can turn into a lot of projects that haven't seen the light of day.

There were also some successes in the indie department. With Eastshade, which was one of the first games I worked on, becoming a sort of indie jewel. Once Upon a Puppet, which was released at the beginning of this year, is a game with a peculiar style, inspired by the art of puppetry. I also worked on projects such as New World: Aeternum and Scavengers.

The Witch of the Waste

The Witch of the Waste is a project from a current series, through which I'm trying to develop a style of my own. In high school, I wasn't in an art-related profile, quite the opposite, in a certain sense. I studied maths and programming, something that was very far away from my interests, passions, and skill set.

So, when I got to the University of Arts to study graphics, I felt I was always a few steps behind my colleagues who had already worked on their artistic capabilities in high school as well. I was also amazed at the ease with which some of them drew. Their hands moved on the canvas as if they were an extension of one another. And, for a long time, this was a goal for me: to reach that ease of movement that very well-prepared artists have when working on their canvas.

A freedom of movement as opposed to the carefully placed dot, the anxiously traced line. After many years, I think I have reached a similar point in my artistic path. My movement has become freer and less tedious. And so, the focus in this series was this: developing a style in 3D through which the visual focus points and details follow the artistic accidents of the hand moving as freely as possible over the digital clay.

I also wanted to focus more on the artistic side of creating digital characters and less on the technical requirements of game art. I'm thinking about this one, and a few others that I finished recently, more as images and artworks that are supposed to achieve a certain feeling, more than characters that have to mark some technical aspects. And I have several reasons for doing this.

The main one is a sort of future-proofing for my job as an artist in the gaming industry. I expect that history will repeat itself. Just like it happened at the end of the 19th century with the appearance of the camera, when artists, to keep their roles in society, were required to focus more on artistic expression rather than realism and technical achievement. I think that's where we might be headed as well, with AI appearing and making its mark in the industry.

That is, I don't think AI will steal our jobs, but it will make them change. Digital art and artists will move further towards expression, style, and substance, and the technical reproduction of things will be taken over by AI, scans, and other technologies.

So, my personal goal with the Witch of the Waste, and other such projects that I did or will do in the future, is to start moving that way: make characters that feel alive and tell stories through minute details, rather than technical excellence with no artistic impact. It's still a process of exploration for me at the moment, but I hope I'll hit the mark soon enough. 

Sculpting

The sculpt was started with nothing in mind, only the goal of a similar style as the previous witch that I have sculpted, also after a concept by Patrick Ganas. In a way, the first step is choosing the right concept.

There's an anxiety of starting from nothing in this sort of work. The artist's fear of the blank page, which translates in 3D as well, into a sphere that looks like nothing. But, when you're a 3D Artist, that fear is sort of mediated by the concept a bit. Because what you have to do, most times, is to translate that into 3D.

So, turning that "nothing-sphere" into something driven by what another artist did to solve the problem of a blank page. From that, the question that arises is: what can you add to make it your own, while also being true to the concept? And, this, I think, is the problem of style for a 3D Artist: How can you make something look distinctively yours when you're already working from the established style of another artist?

I'm not sure I have the answer yet. But that's the problem that I'm trying to solve by working on this kind of project. Outside of this, sculpting is pretty straightforward. I started from a sphere and tried first to achieve some likeness in symmetry, while also adding in a sketched version of all the elements. Here's how a main block-out looked after about three hours of work:

The fur that I had at this point was made with an Insert Mesh brush that I have, which I started adding in symmetrically in some areas, and asymmetrically in areas where symmetry wasn't possible. I started adding color early on to make sure I'm capturing a bit of the feeling of the character.

When working with shapes that are so thoroughly stylized, it can feel uncanny for quite a while there, so just trust the process, as they say, and ignore the anxiety that comes with it. I now find that learning how to ignore the anxiety that comes with the initial stages of an artwork is the first step into achieving that freedom of movement that I was talking about.

After this first pass, I continued by posing the character. Although, in this case, posing it just means having the head turned a bit, you'd be surprised how much personality a simple turn of the head can add to a character. And, with it posed, I started sculpting finer details in asymmetry, while also adding little bits of color to enhance my sculpt.

While doing this, I had a camera angle saved in my ZAppLink Properties, which you can find under Documents, in ZBrush, in order to keep looking at it, from time to time, through the angle from which I knew I wanted to display it. I also changed the ZBrush focal distance to 85, to make sure I'm seeing it as close as possible to the final render composition.

The hair was made with an IMM brush, which I applied a few first strokes of, then started multiplying and deforming manually, to achieve a sort of dynamic, blown by the wind look. While making sure that there are some strokes clearly breaking the silhouette of the hairstyle from every angle.

While arriving at the final stages of this, I also had set up a rendering scene in Marmoset Toolbag. So, with each significant change that I was making on the sculpt, I was updating the model in Marmoset Toolbag, experimenting with lights, angles & whatnot. Unfortunately, I didn't save incrementally on the rendering scene, so I don't have all the stages of how it looked. But I can assure you, it looked quite bad at first. 

In the final stage of the sculpt, and after seeing it in the rendering scene, I decided that the sculpted fur wasn't doing so well with the hair looking so dynamic, so I decided to add some accents of hair strands here and there, to make it more interesting and closer to the concept.

However, after adding some all over the place, I decided to try and duplicate the whole strands tool and move it around so as to give it volume. After doing so a couple of times, I ended up with a mesh that had strands polygrouped separately, and so, to take it further, I started isolating the polygroups one by one, shaping them with the Move brush, and copying them around to populate areas that remained empty.

I also colored the tips and the roots slightly differently to test out how it looks in the rendering scene, and ended up adding a bit of subsurface, with a hint of blue in it, to give it a more rounded out feeling. It's a subtle change, but it makes a lot of difference.

Rendering and Texturing

For the rendering and texturing, I used Marmoset Toolbag and ZBrush in a sort of lax combination. Since Marmoset Toolbag 5 came out, the number of polygons that you can hold easily in a real-time rendering scene with ray tracing on changed drastically. Before, I was always decimating the models that I was importing to preview or render, but now the polycount doesn't matter anymore.

This is a sort of experimental artwork where expressivity and style were the purpose, as I said, so technical requirements weren't something I wanted to bother with. The fewer technical constraints, the more you can do with art. At least that's what I'm hoping we're headed for.

So, I brought in my model in the rendering scene, its textures painted with vertex paint in ZBrush. To make them work in Marmoset Toolbag, you only need to switch the albedo panel to read the texture from vertex color, it's a simple drop-down. On top of that, what I do is fit the tone of the color by adjusting the value from the color picker, in accordance with the light in the scene.

For this one, I set up a two-point lighting system, with the sky left open as a fill light, but on a very, very low value of 0.01. It's basically there just as a parent for the two lights that I added. I used a Spot Light as my main lighting source because I like the control it gives over the softness of the shadow, and the areas that I light with it.

Additionally, I used an omni light for the highlights on the hair and eyes, coming from the front. So, a top spot light for the main light and shadows, and a frontal omni for the highlights and general fill. I wanted a light source to be directed from where the wind would be blowing in her hair. To fake the impression that maybe the light is the one that's doing that, like through magic. Here are two screenshots with the lights separately:

I also always add a bit of color to the light, but it's a lot of trial and error on this. I just go with what seems to work best. I try not to overpower the colors of the textures, but also to drive a feeling through the light.

Finally, the cameras. I like to use separate cameras for each shot that I take. I start by making a main one, on which I determine all the post effects, depth of field, and color grading that I want to use. I also keep, at the same time, a test camera to work on the model, lights and whatnot. After I'm satisfied with the properties of my main camera, I start duplicating and searching for shots in which I want my model to appear.

That can sometimes mean going into the test camera, selecting the active camera that I'm working on, and rotating it manually along different axes to catch a more dynamic or interesting shot. It might also mean moving the lights and the model around a bit more. I'm always open to moving stuff around a bit more once I get my properties right.

Conclusion

Finally, challenges that I had along the road and any advice that I may have can go together. I never settle on the first light setup that I make. I keep all the variations, as you can see, but never go for the first one. I'll change it at the last minute if I find something that seems more suitable. But I always duplicate things when I make a change, so I can easily see the comparison.

I also never settle on the first camera setup. I test things out and make changes in the last minute if necessary. Again, keeping the old versions for quick comparison of the two. Split screen in Marmoset Toolbag is very useful for this. If I do a camera animation, render it out, and after rendering it out, I decide that I don’t like it, I change it, regardless of the hours I spent rendering.

So, if there's any advice that I can give based on how I work, it's this one: never settle for the first thing that you do. Change it up, regardless of how much work you've put into a light setup, a sculpt, a pose, a rendering, a whatever. I think that's what makes the work better each time, even if only just a little bit. It's worth it.

Daniel Merticariu, 3D Character Artist

Interview conducted by Emma Collins