Stunning 3D Model of Harlequin Made in Marvelous Designer & Maya

Introduction

Hello! My name is Tristan Lee, I am a 3D Character Artist, currently looking for work. While 3D is my professional focus in art, I have always loved to draw and paint on the side. I think having multiple creative channels helps me not lose interest in any of them.

I started working in the games industry in 2014 when I attended the newly created Game Design program at Sheridan College. There, I taught myself 3D art because I was interested in it and believed it would be a good career path for me.

After graduating, I worked as a generalist at various indie studios around Toronto, gaining experience and skills. One of my favorite indie projects was a game called Tidal Shock, on which I handled almost everything related to 3D assets. It was on this project that I realized I wanted to specialize in character art.

After that, I was lucky enough to work as a Character Artist at Snowball Studios on a Fantasy Netflix show. This experience was like a Character Artist boot camp, and I absolutely loved it. However, I could not escape the nagging feeling that my portfolio no longer represented my current skill set. Because of that, I decided to do a mentorship term at the Think Tank Training Centre to create a portfolio I could be proud of.

Inspiration & References

I spent about a month on the side deciding what concept I wanted to do during my mentorship. By the end, I had an ArtStation Collection with around 80 projects. I ended up settling on Hocheol Ryu's Harlequin concept for two reasons. First, it was the most striking and loud among the concepts that I saved. I could keep looking at it and noticing more details, which I wanted to capture. Second, it had all the elements I wanted to improve on, such as a lot of flowing cloth and hair, intricate prop sculpts, and a character that felt like it told a story. 

I also love art that exists in historical fantasy settings. There's so much inspiration to be drawn from old designs and clothing, and the fantasy setting adds a little bit of sauce on top.

I know many people recommend gathering all your references at the start, but my ADHD brain doesn't like working that way. During blockout, I typically get one or two images for all elements. After blockout, I pick away at each element one by one. Once I pick one element, I deeply dive into references for that piece to understand how it works. 

Modeling

I started with a quick blockout. Recently, I have settled on breaking my blockouts into two stages. In the initial blockout stage, I focus mainly on getting all the elements in there as quickly as possible (it usually doesn't look great). This means getting rough representations of each element without worrying about proportions. In stage two of blockout, I can focus solely on proportions and the relationship between elements. At least for me, this kind of step-by-step problem-solving allows me to better focus on each problem.

After the blockout stage, I pick one equally exciting and challenging element to sculpt to a high level of polish. This allows me to have something to look forward to after each individual piece and keeps me excited throughout the project. 

A piecemeal approach also allows me to group up similar problems. For example, I can do hard-surface sculpts one after another, then proceed to anatomy sculpts, and after that, I can approach any cloth and Marvelous work as a group following that. Staying in one problem-solving mindset gives me space to experiment within that field and come up with interesting solutions, as opposed to working in several brain spaces at once. I then follow this process for every piece of the blockout, slowly working my way to a completely high poly character.

As I work through the high poly, I choose which pieces I want to start with good topology to save time in retopology later. Usually, I weigh this by trying to extrapolate how much sculpting on top of the final topology with subdivides would hinder my sculpting abilities.

Probably the scariest and most challenging piece was the character's cape. The challenge here was to get an accurate fabric simulation while also matching the shape found in the concept. The method that I and my mentor settled on included: 
1. Start with a cloth simulation in Marvelous Designer.
2. Pin the basic shape with a wind controller blowing the fabric around.
3. Take this into ZBrush and sculpt it into the shape of the concept.
4. Take it back into Marvelous Designer, place more pins, and simulate again.
Once the basic shape looked right, I did all the final folds and shapes in ZBrush, the simulation informing my points of tension.

Something that helped me here is that I did most of the cape work after the character was posed, textured, and placed in a lighting environment. As you can see, I was able to bring each iteration of the cape into my scene with some basic Marmoset materials. It made LookDev much easier than guessing what it would look like once textured and lit.

I think I did about ten different face sculpts for her face until I settled on one I liked. I then wrapped a VFace onto my sculpt and applied the pores in a displacement layer. Once there is realistic skin detail on my sculpt, it's easier to spot details that lead into the uncanny valley. So, I usually work under the detail layer and add fat, wrinkles, and other necessary details to make it all work together.

One final tip I would give for the high poly stage is if you ever reach a point where you feel stuck or slow, take half a day and poly paint your model in ZBrush. It's a totally different mindset, and getting a sneak peek at what is to come in the texturing stage can be very reinvigorating. 

Retopology & Unwrapping

I usually do most of my retopology using Maya's Quad Draw tool. From the blockout phase, I had a bunch of assets that all had relatively the same topology. This is because during blockout I either ZModeled or modeled these pieces with the final topology in mind. A good chunk of her clothing and body I then manually retopologized using Maya's Quadraw. 

The most challenging retopology work was definitely on the white poofs on her legs and right arm. For her bloomers, I followed this method from Paradoox on YouTube:

To sum up, you need to export two versions of your garment from Marvelous Designer. The first is dense and in the pose you want, and the second has low density and is in flat pattern mode. When imported into Maya, both models will share UVs from the garment pattern. You can then delete the internal polys on the flat garment and create a clean topology using these vertices. After completing each piece, you can then transfer the vertex position onto your low from the high based on UVs.

This method worked excellently for the bloomers, but on the poof on her arm had too many complex folds. This meant that my clean, even geo was not capturing the silhouette well. To go in and fix it would be a start from scratch. For this, I used ZRemesh. First, I polygrouped all the areas that I wanted to have clean edges for my UVs. I then ZRemeshed with keeping groups on. This got me about 90 percent of the way there and required a bit of manual work on top.

When it comes to unwrapping, I like to unwrap in Maya as I am retopologizing. This way, once I have all the pieces in, all I have to worry about is UV packing.

Texturing

Texturing is by far my favorite part of a character model. It is then that I see life finally born into this thing I have been working so hard on. 

For her texturing, I wanted to try something a little bit different. I knew that I wanted to do these incredibly close-up shots of all her little bits of jewelry and accessories to show them off. Because of this, I needed to capture a ton of detail in each piece. To do this, when UVing the model, I scaled up individual pieces of the model that I wanted to be highly detailed. Then, when unifying texel density in Maya, it automatically gave those pieces a higher density than everything else. While this may not be a great solution to a game character if optimization is a big concern, I think it makes those close-up shots shine. 

As the big pieces share UVs with the posed pieces, I can also use texture on the large model and hone in on the details.

When it comes to texturing, I like to break up each material into roughly three categories: base material, color variation, roughness variation, and height/normal variation. When I started learning Substance 3D Painter, everything had a disappointing default Painter feel. I have found a way to break out of this. Whether you start from scratch or use a smart material, you need to focus on each of those variation layers and base all your choices on material references.

Here is a breakdown of the cape and hood base material I made. There are base materials that contain roughness, color, and height data. Atop that, I create micro folds using the Crystal 2 texture. A little bit of blur and Crystal 2 creates beautiful micro folds. Then, there is a layer of color variation using the AO map as a mask to create a blue tint within the darker areas of the mesh. After that, there are several layers of color variation with different Fill Layer masks. Then, a couple more layers of height variation will be used to flesh out the texture of the fabric.

Every material follows these basic principles with obvious variation per the asset's needs. I do think the key is variation – in all your Albedo, Roughness, and Height/normals. But I always need to keep in mind which areas I want to draw more attention to, as too much detail in everything will create visual noise.

For her face, I wrapped a VFace at my sculpt and baked the pores into the sculpt. When baking in Marmoset Toolbag, I also baked down the following textures from the VFace onto my low poly: Cavity Map, Albedo Map, and Pore Map. These maps will be very useful in Painter to use as different masks.

I used these maps as masks on the white makeup layer to remove the makeup from cavities and pores. Then above that, I masked the makeup using streaks to create a brushed look.

Rendering, Lighting & Post-Production

I did my final renders in Marmoset Toolbag 4. Using ray tracing, I was able to fine-tune the lighting and shaders to capture a detailed painterly look.

When it comes to setting up lighting, I think less is more. I have been in scenarios where I had so many lights that I couldn't keep track of them all. This makes broad sweeping changes hard. A basic film setup helps me keep my lights limited and it looks more cinematic. The key light highlights the main forms and shadows, the backlight highlights one side from the back, and the fill light is for the darker shadowed areas. In Marmoset Toolbag, I also use the Skylight to assist in my fill lighting and to get some interesting HDRI lighting as well.

A tip that helped me set up my key and backlight is to isolate them and work on each individually. This lets you narrow in on interesting shadows and see exactly what you are highlighting.

I rendered all previous work in Marmoset Toolbag and used default material settings, but there is much more I could have done. In this project, I wanted to capture the feel of each material to the best of my ability. For me, this meant fine-tuning each shader based on the material reference. 

A good example is the inner gold of her cape. I interpreted the cape as two materials: a thick red velvet on the back and sheer, thin gold fabric on the inside. In the shader, I gave the gold material some fuzz and a reddish subsurface glow to hint at the red cloth beneath shining through in the light. I also added detail normals to some of the fabric to create high-resolution tiling cloth details. I think a little extra shader work can make your texture work shine.

Conclusion

In total, I worked on this character for around five months. Because I treated this as a learning experience, I wanted to put as much time into each challenge as possible and become a better artist.

I think the most significant challenges along the way were the mental hurdles, the technical issues, and life getting in the way. While working on the character, I moved twice, once to a different country, and left my freelance gig. I had to put my head down and believe that I could finish this project, no matter what was going on around me. But I faced a lot of mental battles. I stopped working for eight months to return to school, which was scary. There was much self-doubt and a heavy weight put on completing this project. In the end, I am super glad I put in this time and took it one step at a time without letting the stress consume me. I can't believe how much I learned in a very short period of time.

If I could go back and give myself or anyone wanting to become a Character Artist a piece of advice, I would focus on letting go of perfectionism. It was at the Think Tank Training Centre that I learned of the concept of plus-one-ing, which is continuously making tiny little changes over time until you start running out of room for improvement. Previously, I found my worth as an artist in being able to make something great quickly and on the first try. But because of this, I wasn't letting myself learn. I do believe you have to make the mistakes yourself to understand the lesson, but you can't do that if you're too scared to make any mistakes. 

This project didn't look great at the start, my blockout honestly looked relatively bad. But every time I looked at it, I saw something I wanted to fix. Over time, after hundreds of fixes and lessons learned, I looked back at what I started with and could not believe how far it had come.

Tristan Lee, 3D Character Artist

Interview conducted by Gloria Levine