Recreating Across the Spider-Verse's Gwen Stacy in ZBrush & Marmoset Toolbag

Introduction

I am Clément Néca, a Freelance Character Artist, working in the game industry. I am passionate about video games, and I am an anatomy nerd. For me, it was obvious to make characters for games. I even plan to start my own game in a year or two. 

I studied for five years in an animation and game art school in Paris called New 3Dge, where I learned all the basics needed to make a start in the industry. After my Master's degree, I have been focused on stylized characters, and I have worked on a bunch of personal projects.

References

The project that taught me the most was Fang. It is about concept art, stylization, hand painting, etc. When I saw Arcane, I knew I needed a project at this level. Thanks to this project, I was able to feel comfortable with stylized characters and finish the Gwen Stacy project in a week.

When I saw Spider-Man: Across the Spider-Verse, I had the same feeling I had after Arcane. This film is exceptional, so I chose to recreate Gwen and Miles. Like any good project, I collected references on everything that could be useful to me, from the general things such as proportions, silhouette, design, and colors to the specific ones like shading, facial lines, and close-ups. Here are some references I took from trailers and videos:

As there was no concept art, I only relied on screenshots from the trailer and short clips on YouTube. I found out as much as I could about everything that was available, and I took a screenshot of each thing I learned. Then, I put it all into PureRef. This research and analysis stage took me a full day, about 10-12 hours.

This step is often overlooked, and the result inevitably suffers the consequences. Just by looking at the references and thinking things through, you can already have a fairly precise idea of the process, even if you've never done it before. So you save a lot of time and improve the quality of your work. Don't underestimate the importance of references. You learn a lot from them. Let's take an example with a small bit from the references:

Workflow

Now that we've learned a lot from our research, we can imagine what the process will be like, what will be different from what we're used to, and how we should approach these changes. It's best to avoid starting a project without knowing where you're going. Once you have an idea of the process, you can finally start blocking. 

It doesn't matter what software or method you use. You can start with a Basemesh, poly-modeling, etc. The aim here is just to have a silhouette that works. For a stylized project, and especially for Gwen, the silhouette has to be dynamic, so we look at the references and understand the relationship between the curved and straight parts. It's the same approach as in 2D. 

In the beginning, we ignore details, anatomy, etc. We just want a silhouette that works from any point of view. There are always tools to cheat, such as ZBrush's See-through, or you can simply import the reference, but be careful with that. A slight rotation can compromise the understanding of volumes, and perspective can also vary a lot depending on the shot.

Tools

I only used ZBrush for blocking, sculpting, and even painting. In ZBrush, there are no secrets, and I use very few options. It's mainly sculpting with brushes, Move, Dam Standard, ClayBuildup, Trim Dynamic, and Paint. I have my own version of ClayBuildup, Dam Standard, and Paint brushes.

For the ClayBuildup and Dam Standard, I just went into Stroke, Lazy Mouse, and reduced the Lazy Step very low. Then I reduced the intensity to around 12 or 13, and finally, for the ClayBuildup, I activated the BackFace mask too. The result is a cleaner brush. I also like to use the BackFace mask on the move to create volume in the blocking. To avoid having to activate and deactivate it all the time, I activate it on the topological move and deactivate it on the classic move.

Assign keyboard shortcuts to your main brushes. Remember: Ctrl+Alt+Click on the brush to assign a keyboard shortcut. Otherwise, apart from the modified brushes, I just sculpted, staying low poly and avoiding zooming in too much so as not to linger on details.

Modeling

The head is often a tricky part. Capturing a unique style is always the biggest challenge for most people, but it's also what makes the project interesting. It's a lot of observation, trying to understand volumes from as many angles as possible, and applying those volumes by constantly comparing with references. It's 'just' comparing your model with the references hundreds of times. The most important thing, as always, is the silhouette. Thanks to the silhouette and the relationship between light and shadow, we understand the volumes of the face.

You start with the biggest shapes, the length and width of the head, and compare it with the rest of the body. Once you've got that right, you can zoom in a little more and start again to compare the volumes of your model with those you understand from analyzing the references. Repeat the operation, going into more and more detail, but don't go too fast, because if the first steps aren't right, the rest won't be either. You can have the perfect nose shape, but if it's too small in relation to the face or badly placed, it's useless. 

Here are the main steps:

1. Basemesh.
2. Head shape.
3. High and base painting.
4. Final painting.
5. Hair.

For the eyes, it's always best to place the eyeball before doing the eyelids to get the right volume, but sometimes, the eyeballs aren't perfectly round depending on the artistic direction.

And finally, for the hair, I really don't have any secrets. I think there's still plenty of room for improvement. I do 3 or 4 strands that I duplicate and deform one by one to lay the blocking. Then, for the important strands in the silhouette, I sculpt them one by one to make sure I get exactly what I want.

Once the blocking of the hair and the important strands are done, I sculpt all the visible strands to make them unique and try to give a flow to the haircut. And to finish, I place small flyovers by hand, those little hairs that come out of the main shape of the haircut. When working on the hair, always remember to look at it from several angles. Getting an interesting silhouette from all angles is very complicated and time-consuming, but the result is worth it. 

Costume Design

For the costume, it's the same process as we saw before. We focus on the silhouette. It should work from all angles and be dynamic, especially for a character like Gwen, who will often be in dynamic poses and turned in all directions. I wanted this project to be quick, so I started from a Basemesh I'd made for my character Fang, so I already had a slightly stylized shape and a clean topology. But if you start from scratch, try to have a clean topology anyway, with a ZRemesher, for example, so the mesh will be a lot easier to work with than a mesh with a bad and very stretched topology.

For Gwen, we need a fairly uniform topology for the paint we're going to see next. For the hood, I polymodeled it directly in ZBrush with the ZModeler and subdivided and sculpted it to get a more interesting flow in the folds rather than something symmetrical.

They are very simple, so I wasn't bothered by small details. For simple meshes without too much detail, try to stay zoomed out as much as possible to see the character as a whole and not just focus on a specific part that might not work well with the rest.

Texturing

For the retopology, as I said above, I started with a Basemesh for the body and for the head. For the rest, I made it in polymodeling to have a level 1 subdivision that I could use as a low poly. I often do this because it saves me a lot of time when I'm doing the retopology. 

The other big advantage is that when you've done your UVs, you can re-import the objects at subdivision level one, and it will keep all your subdivisions, but update your UVs. So if you need to touch up the sculpt later, you don't have to readapt the low poly every time, and what's more, it will be very useful for the paint job. 

However, be careful, using subdivision level 1 as your final low poly doesn't work if, in your retopology, you need to merge as many objects as possible into a single mesh. For example, a belt that would be retopologized in the same object as the body. This is often the case with game characters. For Gwen, the only object I had to apply the retopology was the shoes. The rest already had a good topology in level 1 of the subdivision.

For the texturing, I had a lot of questions from many artists, so some people will be disappointed, but there's no any magic formula. It's literally just paint.

Apart from the costume’s gloss, everything else is hand-painted. I painted everything directly in ZBrush. It's not the most practical, I'm aware of that, but I like the feeling and the constraint of not having any layers. It encourages you to be sure of what you're doing and as I did oil painting at school, and continued on the side for several years, I've got a bit of that feeling of having everything in a single layer. But the software you use isn’t important. Substance 3D Painter or 3D Coat for example are good too.

I'll explain the process and the tools used for those who want to try it out, but I don't think there's anything innovative about it. I started by doing the UVs and importing the mesh with UVs. Then I applied a material I've created, which is a variant of Skin Shade. I've just changed a few parameters, which I'll put here: 

I used my Paint brush, which I've also modified from the basic Paint brush. I've given it a square alpha the same as the ClayBuildup and reduced the focal shift to about -60. I also set the LazyStep to a very low value, as with the ClayBuildup and Dam Standard. If you don't want to make them, all the brushes are available for free here.

So, once my shader has been applied, I check that I'm at a high subdivision level to have a resolution that I like, and I fill an object with a base color. In my case, I'm colorblind, so I pick my base color directly from a screenshot with fairly neutral lighting. Once I've got my base color, most of the time I start painting and work in the same way as I do with oil paints and sculptures. I start with the most important part and then work out the details little by little, gradually zooming in on my model. There's no secret to it. It's all about time, observation, and interpretation. But it's also the best part of the project.

If I don't have a lot of time, I'll try to have a base of light and shadow on my colors to bring out the volumes. Be careful, it must be light so that it doesn't interfere with the lighting of the scene. To do this, there's a little tab in Masking called Mask by Normals. I place myself in the viewport as if I were the spotlight that's going to light up my character in the paint and I click on Mask by Normals. Then I play a bit with my mask to boost it or make it sharper depending on what I want, and I do a Fill Object once again with a darker or lighter color depending on the mask you have. This will be my base color for the shadow or light.

There's also Mask by Occlusion which can be useful, but be careful, depending on the mesh it can take a long time. Once the base is done, we are at the stage I described earlier, painting. No magic. I take my time to observe, analyze, interpret, and replicate. 

If you really need to make a big change when the paint is very advanced, you can always export the map (ZPlugin, Multi Map Exporter, Textures from Polypaint) and make the changes in Photoshop, so you can re-import it in ZBrush after, and Polypaint by texture. That's why having a low poly with clean UVs in ZBrush is useful, but I admit it's not the most practical. The best thing is to be sure that each step is right before moving on to the next, but that's easier said than done. It's something that will come with time and experience. You can't imagine how many failed projects there were before Fang came to life.  

It's the same process for eyes, hair, and so on: base color, light and shade, color variations, imperfections, brushstrokes, highlights, etc. Each element will be approached differently depending on the stylization of the character. That's why I love studying such incredible artistic styles as Arcane or the Spider-verse films.

Rendering

Once we have our mesh ready, we can render it. Marmoset Toolbag is a good solution when you want to quickly render a game-ready mesh. For the Turntable, the render is pretty straightforward. I just put 3 or 4 lights, one main light, one fill light, and one ambient light. And for Gwen, the rim light fit well with the style, so I also put a strong rim light.

For the beauty shot, this is another story. I did some tests on Photoshop with a very quick render, just to try things with colors and composition. That’s how I ended up with the comic book idea for the background. So after studying some comic book covers, I finished the background. After that, I needed to render Gwen too, so she can fit with it.

To do so, I added a background in Marmoset Toolbag and imported mine. I created a camera for the render and a camera for moving into the scene. To make it simpler, I split the viewport in half so I always have a viewport to move in the scene and another to constantly have the final render, with the background.

Since the background is very flashy, I don’t want Gwen to be lit the same way because it will look weird. So I decided to make her pop up from the background with a strong rim light like we have in a lot of comic book covers. And that’s basically all. I just take care that every light looks good alone, and I remove all effects on the camera because I want to have full control of these effects in Photoshop later on.

Post-Production

Now it’s time for the post process! And that is a very important thing in a Spider-verse style. It’s always the same thing. I look at screenshots; I analyze how they stylized the lights, the shadows, the colors, the effects on the sides etc, and I apply what works with my character.

This is how I ended up with hatches in the shadows, little dots to make the transition of the lights, chromatic aberration, etc. By the way, every effect is handmade. I’m pretty bad with Photoshop tools, so I created all the patterns I needed and placed everything by hand. Probably not the best way of working, but I like to have full control of what I do.

And the last part is the feedback. I’m lucky enough to have great friends in the game industry, so I asked them what they think of it, and they spot things I don't always see. 

Conclusion

The entire project took me a week, but I started with a Basemesh for the head and the body (the suit), and I also had a base for the hairs.

People imagine that it's capturing the style that is the most difficult, but I had to do it a lot these past months with the Fang project and with my last job at Quantic Dream. So it wasn't that hard. The challenge here was the post-process. I'm 24, but I'm an old-school artist, so I don't use Photoshop brushes or effects, and it was hard to do exactly what I had in mind on a project that needed a lot of effects.

I'm still pretty young, even if I started early, so I don't feel comfortable giving advice to Character Artists, but I would say not to be afraid to fail. Take your time to find solutions to your problems, do things right, and focus on quality instead of quantity. 

Clément Néca, Character Artist

Interview conducted by Theodore McKenzie