Using Blender's Grooming System to Create Bizarre Jar-Like Dwarf

Introduction

Hello, my name is Juan Hernández and I am a 3D visual development artist. I am self-taught, started as a hobbyist at the end of 2017 and quickly got into the industry. I love making characters and creatures in 3D, both realistic and stylized. Since I started with 3D, I have always been interested in learning all I can about the whole process of creating something for film and animation, from sketching and ideation all the way through asset build until rendering.

The Dwarf Project

One day, out of nowhere, YouTube suggested a video from a channel called ‘Monstergarden’. It was a 5-minute clip made by Simon Lenz showcasing this character and his thought process when creating it. I really loved the final result and I liked how everything in his process felt very grounded and well developed. I also really enjoyed his idea of moving away from the ‘human being with slight modification’ thing and instead keeping the key component that made the character recognizable. From a physical point of view, that would be his big nose and beard, then there are other elements associated with dwarves, like working with metals, living in caves, and so on.

His design was very inspiring, it felt fresh and creative, so I thought developing it further could be a lot of fun. I also really liked the whole mechanics of it being able to change from a humanoid figure to a jar shape. I told myself that it could be quite hard to get right and potentially a total disaster… so, naturally, I was like, "Let's do it!"

Before starting the project, I did some reference gathering. In this particular case, the number of references was on the lower side since I already had a clear design and quite a lot of information about the character from the creator.

Modeling

Making sure the dwarf could transform into a jar meant that blocking out the character properly was one of the most important steps in the whole creation process. Proportions and primary shapes had to be right so the character was not only able to transform but also to ensure it looked good in both poses.

To achieve this, I first blocked out the character in Blender in its final jar pose using simple geometry and moved stuff around until I had a clear jar shape from all angles while including all the elements of the body. 

Once happy, I made a very simple Blender rig and used that to unfold the character into a standing pose. Blender has this interesting concept where you can alter your rig in Edit Space, which happens pre-deformation. This let me change and tweak pivots and bone orientation and instantly see the final pose without messing up the original mesh. This way I check how the legs and arms should rotate and in which direction.

As you can see, now I have a standing pose that can transform the dwarf into a jar shape. If you look carefully, you will notice that the character looks like total crap. Now it was time to make him look better in his standing pose.

One thing to consider is that I should be careful with the changes I make in this pose so that it does not break the fit I did for the jar shape. Luckily, because I am using a rig to unfold the character, I could easily go back to the jar pose and check everything is okay. So I set keyframes for one pose, switched to the next frame, and set keyframes for the other. This lets me toggle between standing and jar poses with one button. From there, I started improving the character and adding a bit more detail while always toggling poses and finding a middle ground for changes. This probably took the longest since I had to sculpt and make big changes until everything worked for both poses.

Once I was happy with the result, I took this geo to ZBrush where I added more shape and logic to the body. At this point, I knew that I would be building on proportions that work and even if I did some small alterations, I would still be very close to what I needed for the transformation, and any tweaks at the end would be small.

To make the patterns, I used this neat and relatively new tool from Substance 3D Painter that lets you create 3D splines that snap to an object’s surface. These work similarly to paths in Adobe Illustrator or just curves in any normal 3D software. The key thing is that you can tweak its thickness and the type stroke so even after I laid out all the patterns, I could play around with the look of the lines. This was very helpful because I could try many things quickly without having to actually sculpt them. Also, it meant that all the patterns had a consistent thickness and profile, which could be hard to achieve by sculpting manually. Eventually, once I was happy with it, I exported the final result as a displacement map and applied it in ZBrush where I could still do a pass of sculpt by hand, to make it look more organic.

Beard

All the groom was done in Blender. Blender has a relatively new grooming system, all node-based and very powerful. Additionally, I tried out an add-on called ‘Medusa Nodes’, developed by Irakli Kupunia. Medusa nodes introduce a custom workflow on how to build grooms using Blender’s existing hair tools (a workflow that I am more familiar with, for this reason, I thought it would be an excellent opportunity to try out). I have been following the work of Irakli since a long time ago when he was experimenting with hair early on even before there were official hair nodes released.

I started by adding a couple of curves that define the actual directionality and length of the beard. Then on top of that, I can start layering the elements like clumping, curling, and flyaways to achieve the look that I want. I made a first version to test the whole workflow, and it ended up a bit like a hipster barista. But then I tried some better shapes, clumps, and messiness, as well as a proper dwarven braided mustache.

For rendering I used the Principled Hair shader and just added a bunch of color variation to make it look more interesting. You can read attributes from your groom in your shader with these 'textures’. Stuff like darker roots, color variation, random grey hairs, and roughness variation can bring your groom to life.

Texturing

I do retopology in Blender, by hand, sort of old school I guess. But Blender, being very shortcut-based, lets me add geometry very quickly with a lot of control. I usually do retopology so I have the flow that I want, but knowing I will add one subdivision level later. That way I know I will have good flow and quite a lot of detail without having to create a dense mesh manually.

UVs are also done in Blender, I have a mouse with two additional buttons on a side and use them in a couple of ways that speed up my work. One of them is making or removing UV seams. That way I can do all UV prep instantly as I select things without having to open any menus or jumping around my viewport or my keyboard. Additionally, I try to take advantage of the mirror modifier as much as possible. One use is to offset UVs of the right side of the model, that way I only have to maintain one. That makes model updates and maintenance more flexible.

All texturing was done in Substance 3D Painter. For the shell, I wanted to achieve a metal look but without it looking like shiny armor. For this reason, I made the surface generally rougher and used skin-like high-frequency patterns to reveal glossy bits. This makes the overall surface look more metallic and interesting in light without it being a fully shiny thing. Simon explains that the patterns on his body are something like ‘tattoos’ and they are carved on their own bodies as ornaments. Thinking about this I made some edges of the patterns chipped off and more glossy. 

One tricky bit was making the shell and the inner body better fit together visually. The orange color really stood out and this needed a couple of iterations until looking right. I ended up using dark gradients towards the limbs and in areas connected to the shell to make the connection more pleasing to the eye. Also added a bit of goo in these areas, to hint at some sweat in tight areas.

Rendering

I render everything in Blender's Cycles. The only difference from the default configuration you get with vanilla Blender is that I am working in ACES color space. Therefore, my personal pipeline is set to ACES, from texturing to actual rendering for post-processing.

I don't really enjoy making realistic environments, if that's what I need, then I use any type of available resources that will let me present my characters the way I want to. Generally, it is a mix of Megascans and assets I've made in the past, then I usually play around with the shaders so I get the correct colors and values for my final render. I can be very lazy with this… for instance, the background of this render is actually just a photograph from Adobe Stock with the character and some CG elements on top. I thought if filmmakers do the same, then why not?

In terms of ideas or composition, I like to doodle around on a paper or to do some reference gathering to get some ideas. For this project, I looked around at nature photography and picked stuff that could potentially work for my presentation, and after that built CG scenes for rendering. 

Eventually, everything is rendered as .EXRs and ran through DaVinci Resolve where I added some basic stuff, like grain, lens distortion, and basic color correction. Generally, I just want to add some elements you get from using an actual camera.

Conclusion

I did write down the time I spent on it in order to track it. Altogether it was about 3 weeks, including preparing the presentation. One big challenge was definitely the animation. I am aware it has its issues, but I am no animator so I try to loosen up a bit about it. I am happy I get to present an idea and hopefully soon get better at it.

To those getting into 3D, I would say try not to spend too much time worrying about software and hardware and rather spend it doing what you enjoy. Train your eyes with the stuff you see online so you are able to give yourself notes objectively when working. And finally, do know that 3D art is all about iterations. There is no such thing as going from start to finish in one go with no mistakes. If something you work on does not look the way you want it to, you can always improve it.

Juan Hernández, Visual Development Artist

Interview conducted by Theodore McKenzie