Coming up with the concept

Pierre-Antoine Moelo: When Damien Bousseau approached me with the idea of creating a reference image to create a collection of stylized materials for Substance 3D Assets, I decided to go back to a series of islands I had created back in 2014. These micro-worlds were a decisive moment in my art journey and I was keen on seeing where I could take the concept.

When I begin a project, I usually go in many directions. Sketching is a great way to quickly explore and evaluate ideas — mine and the client’s.

I prefer to sketch on paper for the tactile aspect: instinctively, this is the way I feel the most comfortable. The technical constraints of paper (I can’t zoom) prevent me from getting lost in the details, so I don’t get too attached to my sketches and I can quickly get to the next one.

For Bell Island, we wanted to have a richness and a large diversity of materials inside the micro-worlds. This was an important aspect of the production from the beginning since we were going to produce all these materials and add them to the Substance 3D Assets platform. So we set out to create a viable micro-world, rich enough to represent a self-contained adventure. There’s a contemplative aspect in small worlds that resonate with me; they are like a poem you can explore and take time to digest.

As a child, I played a lot with Mighty Max toys, which bring together the macro and micro scales, which I find super-interesting. Still, as an adult, I’m very inspired by the paradoxical size of the object as opposed to the world it contains — I’m always drawn toward miniature objects and stop motion films.

Quickly, the story of Bell Island started to emerge from the sketches.

After these drafts made in Photoshop, we worked together with the team to home in on what would work best. It was important to decide what the size of that universe would be, bearing in mind we were in the context of a production and needed to worry about time and budget.

One of the concepts the team liked best at first was the B concept. Originally, it wasn’t even about a bell maker: it was just going to be a bell ringer. The team liked the bell, but also the elements from other island concepts, like pipes, wires, balloons, and so on. So I decided to enrich the B concept with those elements.

This final concept is basically a Frankenstein’s monster of all the earlier ideas. All in all, it took several weeks to slowly enrich the first sketch into what became Bell Island.

As we were circling around the island concept, I also worked on props to define the DNA of that little world we were creating. It was a way to set down the basics of the look and feel of the island; it was important to be as close as possible to the stylized 3D result so the content team would have a solid level of details. I naturally focused on the bells, of course, but also ceramics, as well as patterns.

Getting from a 2D concept to 3D assets

At that time, the team joined in and started the first drafts of materials. To support the team’s production, I had created a style book. This is a reference document which codifies the art direction and makes sure that all the artists have the same guidelines; in the context of a stylized production it helps to have this anchor into what are the main principles of the project’s style.

This is what it looks like:

The entire style rests on 3 big design principles:

Richness of scale in the objects and texture. This principle, called ‘small, medium, big’, is a way to guarantee a diversity of scales in all the elements of the design, from modeling to texturing. In order to design this effect, details are placed carefully, and not at random.

Rough/Coarse scratches and defects over everything. We actually applied little holes and big scratches over every material, like a varnish. From the start, we chose to avoid ultra-detailed elements, so a big part of the project was to stop ourselves from over-rendering models and materials.

after / before

You’ll see this often in stop motion animation because the materials will show scratches from the brushes or other impacts during the model’s construction phase.

The rule of 30%. In order to stylize a realist material and match the art direction, we set a rule to simplify an asset by at least 30%. That meant 30% less grain, less noise, 30% simpler shapes, and so on. Materials and shapes needed to still feel real, but in a cartoonish world.

Bringing together these three principles gets us to mixing the big and small together: with a realistic material, large grain, and rough defects, we end up feeling like the final result is somewhat miniaturized. You can see this in the boat here, which feels small:

Producing the island

Damien Bousseau: We knew we’d have to build around a hundred materials, and around 50 assets to create the island model. At first, I began to create all the filters that reproduce Pierre-Antoine’s painting style, step by step. I isolated the different elements he uses to paint (brushes, layers, lights, etc) and created generators that aimed to reproduce those elements in the order he uses them.

This was an important step for us to understand the way Pierre-Antoine works. And it was essential to get it right so we could emulate his style.

In parallel, the team divided the content production in themes — weed, stone, metals, and so on — and distributed it. 3 artists worked full-time on the project, with support on specific elements by 4 more artists, though everyone participated in the creation of the central elements. We set up a validation system with Pierre-Antoine for each category to make sure we were getting the base right. And matching what Pierre-Antoine saw it in his mind was the whole challenge of this operation. For the team, there was no specific trick, nothing technically difficult in Substance 3D Designer for the construction of the materials. It was all about understanding Pierre-Antoine’s style and being faithful to his direction.

Here is a sneak peek into the construction of the Stylized Coast Cliff Rock Overgrown material:

You can look into the material’s graph yourself! Download it here. Here, we see the different steps we typically took to build a material.

Once we validated one base material — a rock, for instance — it confirmed that we knew how Pierre-Antoine wanted a rock to look, and we could reproduce his artistic style. That allowed us to create new variations: entirely new rocks, in the same spirit. And that’s basically how we approached creating the collection.

https://video.tv.adobe.com/v/3423740

Thanks to these validation steps we were able to create 122 varied and rich materials following Pierre-Antoine’s unique style. They are all available on the Substance 3D Assets platform (you can browse all our stylized assets here).

As always, one major point of the Substance 3D materials is how we approach the asset’s parameters. Here, the way the materials are customizable is entirely part of the art direction. Check out how you can play around with the stylized metal bell material, using Substance 3D Sampler:

https://video.tv.adobe.com/v/3423736

And it’s not just the materials! The atlases are parametric as well to guarantee that they will be able to match your scene exactly. You can choose the presets we have curated for you, or you can play around with the parameters until you get what you want. Or a happy accident.

https://video.tv.adobe.com/v/3423742

Creating a real-time scene

Having all these elements designed, built, ready to use, was an awesome opportunity to stress-test them into a real-time use-case. And so we created the Bell Island video:

https://youtu.be/otPoFapLFpU?si=2rBd2KVTClgu--2Q

We decided to use Unreal Engine 5. Bringing the scene together started with importing the models from the stylized collection — all available on the platform.

After that, we set the materials:

And this is what the final scene looked like.

The scene is pretty big: we’re totaling 31 reference foliages, and 600 meshes, which total up to 250,000 polygons which were then converted in nanites in Unreal Engine 5. With Nanite, there is no more trade-off between visual quality and polygon budget or memory usage, we can focus our energy on creating a beautiful scene, period.

final unlit

The real-time environment was an excellent way to play around with our scene, letting us change the mood fairly easily, too:

https://video.tv.adobe.com/v/3423741

Parametric clouds

Here’s an aspect of this scene that was a little bit challenging and fun: the clouds.

Initially, the clouds were imported as geometric objects and we iterated through many variations of shaders to render them. However the result was not satisfactory: we could not get the fluffy contour nor the impression of depth and light scattering. So we decided to go for a completely different and more realistic approach: volume rendering. Thanks to an article written by Ryan Bruck on shaderbits.com we knew it was possible.

The first step was to voxelize the cloud meshes, that is to say, transforming the triangle mesh into a pseudo 3D texture where the mesh is transformed into lots of 2D slices. To do that we used MagicalVoxel to generate the slices (a giant 256×65536 texture) then we built a custom tool in Substance 3D Designer to organize the slices into a 4096×4096 texture.

The input voxel mesh and the resulting 3D Texture “atlas”

Switching back and forth between application being time-consuming, we built a couple of other tools in Designer in order to edit, transform and previsualize the voxels. It led to the implementation of a Voxel Render node and a simple node to offset the voxels on the 3 axis.

https://video.tv.adobe.com/v/3423737

Finally, we baked some 3D noises provided by default in Designer (and made new ones) and imported everything into Unreal Engine. There we started from a volumetric shader shipped with Unreal and modified it to fit our needs: blending with 3D noises, density function, lighting, distortion…

https://video.tv.adobe.com/v/3423738

Early test in Unreal Engine

Integration and lighting test in Unreal Engine

Real-time Animation

Bell Island needed to move to feel alive. The animation was based on physics; as it is a small island, it was important to feel the wind’s impact on the objects. Elements move, like the bells, the balloons, the vegetation, as well as myriad other elements.

https://video.tv.adobe.com/v/3423739

The smoke is especially fun in its design: it changes as it moves and then disappears. Just like real smoke dissipates in the air.

https://video.tv.adobe.com/v/3423735

Bell Island was a fun project to work on: from Pierre-Antoine’s design to the creation of assets, right up to building an animated scene: it was a wonderful way to stress-test our models and materials in a real-time context. We hope you enjoyed this project, and if you would like to see other collaborations with other people, let us know!

Damien Bousseau: I would like to warmly thank Pierre-Antoine Moelo for his work, his commitment, and his passion. A special thanks also to Simon le Paih, Stéphane Fontaine, Jean-Bastien Juneau-Rouleau, and Nicolas Wirrmann for their work on the real-time experience.