Breakdown: Nice Stylized 3D Model of Odin

Introduction 

Hi! I'm Thales Simonato, a Brazilian 3D character artist based in France. I started out as a generalist, but pretty quickly gravitated toward texturing and lighting. From there, I got curious about grooming, so I did a personal project called The Blue Monkey to test the waters. It worked out really well, and that confirmed for me that fur and hair were what I wanted to specialize in. I followed it up with a Baby Chewbacca piece, which doubled down on that theory.

After investing several years deep in lookdev (texturing, shading, and grooming), I've recently come full circle to sculpting and modeling, focusing more on design and fundamentals. That shift is exactly what led me to the Odin project.

On the studio side, the most notable project I've contributed to is Minions: The Rise of Gru at Illumination Mac Guff. I've also worked on the Netflix short Canvas, among others.

Odin

I discovered Odin on José Manuel Fernández Oli's portfolio, and I was immediately drawn in by the shape language and the design. It also happened to be exactly the kind of concept I had been looking for: a piece with a huge variety of elements all in one image. Metal, leather, skin, beard, feathers, the whole range. I knew it would push me technically, and I decided to take it on as a challenge.

My main goal was to get the design right. After spending so much time deep in Look Dev, I wanted to put fundamentals first this time around. Forms, silhouette, proportions, the things that make a character read before any texture or shader gets applied. Working under the mentorship of Hannah Kang and Travis Bourbeau was a big part of that, since their eye for shape and design pushed me far beyond where I would have gone on my own.

Modeling

Before touching a sculpting tool, I spent a lot of time analyzing the concept art. Straights versus curves, proportions, negative space, how many head heights tall the figure was, where the apexes sat. That breakdown step is what tells me whether I'm chasing the design or just copying surface details.

I also built a dedicated PureRef board with anatomy references: a mix of stylized drawings, realistic studies, photos, and sculptures I admired. My workflow during sculpting was to keep both PureRef boards open side by side: one with the concept analysis, the other with the anatomy. That way, I could pull design decisions from one and check anatomical correctness against the other in real time.

For the sculpt itself, I went from scratch in ZBrush. No base mesh for this one. I wanted the proportions and the gesture to come from the analysis stage, not from an existing topology I might subconsciously defer to.

The beard had a specific plan from the start. I knew I wanted to test a tool called Ahoge, which fills NURBS tubes with hair, so I sculpted the beard already as a series of tubes with the ends opened up. That way, I could feed them directly into Ahoge later for the grooming pass, without having to redo any of the silhouette work. That kind of forward planning is one of my favorite time savers: knowing what tool you're going to use later, and modeling now to match its needs.

To really nail the beard's design, I also made a clown version of it, with each sub-shape painted a different bright color. It's a great way to break the form down into clearly readable shapes. If two pieces blur together visually in that version, you know you need to push the separation in the sculpt before moving on.

The beard also has some flyaways for extra realism and movement. My technique for those was to export the beard to Maya, draw each flyaway using the Curve tool, add a sweep to give it geometry, and then bring it back into ZBrush to integrate with the rest of the sculpt. It's a roundabout path, but it gives me much more control over the placement and curvature of each strand than trying to sculpt them in directly.

For the crows, I sculpted just the base shapes without any surface detail. The fur was all done later in XGen. The crows aren't feathered, by the way, they're covered in fur, which is a stylized take on the character.

The other big time-saving trick was my approach to secondary elements. The crows, the spear, and the cape were all kept as design placeholders for most of the project. They were just rough enough to read in silhouette and composition. Only at the very end, once everything was locked, did I go back and clean them up into final assets. That stopped me from over-investing in details that might end up cut or redesigned along the way.

I should also be honest that I'm not very deep into ZBrush's advanced features. I don't really use IMM brushes, NanoMesh, or instancing. So for elements that needed precise repetition or procedural building, I went outside ZBrush. The plate skirt was done in Maya, where I could lean on simple duplication and array tools for the repeating plates. The base mesh of the braid was built in Houdini. Both could have been done in ZBrush, but I went where I knew I'd move fastest, and that's a time saver in itself: use the tools you actually know well, instead of fighting to learn a new feature in the middle of a project.

UVs

This project was not built for a production pipeline. It was more of a visdev approach, where the priority is getting strong-looking results fast, even if the topology and UVs aren't perfectly clean. So I kept the topology very simple, just enough to serve my purpose, and treated UVs the same way. For retopo, I used Quad Draw in Maya, and for the unwrap, I stayed inside Maya's built-in UV editor, mostly relying on its automatic unwrap with quick manual editing on top.

I worked with a UDIM pipeline to get higher resolution in the textures where it counted, but the goal was always "good enough to render beautifully", not "good enough to hand off to a rigging or texturing team".

You can actually see in the UV layout that some pieces of geometry have no UV unwrap at all, and the beard's UVs are completely stretched. Since I wasn't using that geometry visually in the final render (the beard's surface is hidden under the hair generated by Ahoge, not surface-textured geometry), I just ignored it. It's a good reminder that not every part of a model needs the same level of care. Spending time on UVs you won't use is just wasted hours.

Texturing

For Odin, I decided to work fast on the texturing, since the main goal of this project was always the sculpture and the design. So I leaned on a streamlined visdev approach instead of a full production texturing pass.

Preparing to Texture

The first thing I do is what I call my BasicShader pass. It's a technical step that's a core part of my process. I render with color and specular only, and I try to define the entire look of each material using just those two attributes. This step actually serves two purposes at once. The obvious one is establishing the material direction. The less obvious one is that it doubles as a sculpt validation pass. If the specular is reacting strangely to a surface, that's almost always a sign that the modeling has an issue, not the shader. So I'll go back and fix the sculpt before pushing further on the texture.

Once I'm happy with that color and specular pass, I render an EXR with an ID pass for the different elements and take it into Photoshop. From there, I play with the colors and values until the whole thing reads the way I want it to.

In this case, I created a Photoshop Action to speed up the process. It adds an Exposure adjustment and an HSV adjustment into a group and uses the current selection from the ID pass as a mask on that group. One click and the setup is ready for non-destructive editing.

Once I'm happy with the Photoshop adjustments, I bring everything to Substance 3D Painter and try to apply those values there.

Texturing in Substance 3D Painter 

In Substance 3D Painter, the workflow is pretty straightforward. I start by rebuilding my BasicShader inside the software, using the specular I set up in Maya and the color from my Photoshop pass as the foundation.

Once that base is in place, I start layering details on top: color variation, grunge textures, and scratches. I also lean on Substance 3D Painter's automatic generators for added realism, especially curvature and Ambient Occlusion, which help the surface detail respond naturally to the form.

Grooming

The hair and fur grooming was split across two tools. The beard was done in Ahoge, which is what all the open-ended-tube work in the modeling stage was preparing for. Then XGen handled both the cape and the crows. As I mentioned earlier, the crows aren't feathered; they're fur, which made the XGen setup more consistent across the two elements since they share the same kind of grooming logic.

The most challenging part was the weaving for the pants and the bandages. I had the pleasure of beta-testing a tool that's coming out soon called Fibric, and it's just amazing.

When you work with weaving tools, the hardest thing is getting the scale and the amount of detail right for the character. I didn't want the weave to look fully realistic, but I wanted it to feel believable. Even though the scale is pushed a bit, I think it landed at the right amount of detail. Convincing enough to read, but not so much that it pulls attention away from the rest of the design.

Lighting

I started with a basic studio lighting setup, but with the character already in pose. For this project, I didn't do my look dev in T-pose. Once shaders react to the actual silhouette and the actual angles the character will be seen from, the decisions you make there hold up.

Once the shaders were heading in a good direction, I moved to the artistic lighting pass. The final setup ended up with 15 lights, which is unusual for me. I normally work with very few. But here I wanted very precise control over the volumes, so I went heavier than usual. On top of the lights themselves, I used 5 light blockers and several decay setups to shape the falloff exactly where I wanted it.

Render

The renderer was Arnold, with everything color-managed in ACES CG. To my surprise, the final render came in around 15 minutes, which is fast considering the amount of grooming, the weaving, and the 3K output resolution. That said, there's a real cost to the multi-light EXR workflow I'll describe below: a single frame ended up at 1.1 GB since every light split adds its own layers.

Post-Production

I try to keep the volumes untouched in Photoshop. The lens-style finishing touches like vignette, grain, and chromatic aberration are fine, but I want the actual reads of light and form to come straight out of the render. If a volume looks off, I'd rather fix it at the source than patch it in post.

The exception this time was the beard. Since the beard doesn't have UVs, only a root-to-tip parametrization, I couldn't drive its gradient through textures the way I could on the rest of the character. So I ended up reworking that gradient in Photoshop to give the face better separation and to push the beard's own volumes more clearly. The other small fix was a single hand-painted mask because of a bug in Maya's Lighting Link tool. Outside of those two corrections, everything in the final image came directly from Arnold.

That said, I do use Photoshop as part of the lighting loop itself. I render an EXR with each light split into its own layer, take the file into Photoshop, and rework the intensity and color of each one until I'm happy. Then I bring those adjustments back into Arnold and re-tune the actual light values to match, so the next render comes out of Arnold looking the way I just decided it should.

Conclusion

The biggest challenge for me was developing the eye: reading the right shapes, proportions, and curves. Having all the design fundamentals in mind and actually applying them in the moment, not just recognizing them after the fact. That's the work, and it doesn't really end.

If I had to summarize what I took away from this project, it would be two things. First, fundamentals beat technique. The design has to be read first, before anything else. Second, plan your modeling around the tools you'll use later. The beard tubes for Ahoge are the clearest example, but the same logic applies almost everywhere. Knowing where you're heading lets you save real time at every stage.

Advice is harder. Honestly, I don't think I have THE answer. I believe everyone is already doing their best and following what they think is right for them. But if I had to share what I'm working on myself right now, it would be this: design fundamentals are more important than ever. With AI taking on more of the technical side, knowing why something looks good or bad is what's going to matter in your career. That's exactly what I'm trying to get better at, and learning to draw is a big part of that challenge for me too.

Beyond that: believe in yourself, surround yourself with good friends, be gentle. And get the design fundamentals right.

Thales Simonato, Look Dev Artist

Interview conducted by Gloria Levine