Recreating a Hobbit Hole in 3D

Part I: Alexius Hill

Introduction

Hi, my name is Alexius Hill. I am a student at the University of Colorado Denver. I am studying as a 3D generalist, but on my own time, I am specializing in Character Modeling. I completed a BFA at Mercer University in Graphic Design and I received a job at a Graphic Design studio in Denver, but I quickly realized that Graphic Design was not the path for me. At the time, I was going to school and working as a Graphic Designer. I had to make an extremely tough decision, to either pursue my newfound passion for 3D or split my time between the two. I chose to quit my job and move back in with my parents. Since then I have been teaching myself as much as I can, absorbing all the information that I am given, and collaborating with some amazingly talented students at my school who have been instrumental in my learning.

Hobbit Hole: How It Started

For the Hobbit Hole, Noah Kentwortz and I were teamed up by chance in our environments class at CU Denver. As our first assignment, we were put together in a group and tasked with making a Lord of The Rings inspired piece, the only requirements of which were the subject and the camera move. We started August 17 and we knew our deadline was September 21 (about 5 weeks) giving us about 3 weeks of work time. We were drawn toward the Hobbit Hole because of the appeal it had. In particular, we were drawn towards the lighting, the foliage, and the small contained scope considering we only had a two-person group.

Reference

Once we decided what we were going to do we each made a reference board. The top photograph of the hobbit hole is a mash of all the elements that I liked about each house. After I identified what I liked, it was much easier to start breaking it down.

During my reference gathering phase, I stumbled upon ​Ivanna Liitschwager’s tutorial, “Sculpting Environments for Games”. In her videos, she also sculpted a Hobbit Hole. It was extremely helpful to see how she started her process and the tools she used like brushes and alphas to accomplish the finished product. I took her advice and adapted it to fit my style, the vision for the project, and our needs as a group.

Blockout

I did my initial blockout in Maya. The goal with this blockout was to get the right forms and keep everything as low poly as possible because I planned on sculpting in ZBrush. I wanted to have several subdivision levels to refer back to because I would be baking my HP to my LP in Substance Painter. I did UVs at this stage but I realized later on, after I finished sculpting, that I didn’t need to do them. The model changed so much after the sculpting phase that UV seams had to be hidden in a different way.

Sculpting/Modeling

Everything in this scene was sculpted except for the lantern and the windows. I started with the wood and worked my way around to the rock and the stucco. I tried to work on objects of similar texture. I found that by doing it this way it was easier for me to get into a flow. Due to time constraints, I did not make my own alphas for the wood or the rocks. Fredo on Gumroad and ​Pablo Munoz Gomez​ on Artstation have some great brushes and alphas. Definitely check them out!

For both the wood and the rock, trimming down the edges with the trim dynamic brush is a must. It informs the wood and the rock and allows you to find the shape. Then you can go in with different brushes such as the Damien Standard, Trim Smooth Border, Clay Tubes (with color spray), and the Orb Cracks brush to really accentuate the subtle details that you made. Sometimes it helps to lay down an alpha and then blend and add more detail from there.

Exporting Phase

Before I could jump into texturing I needed to optimize the door, the rocks, and the bricks. I formulated a technique that I usually use for retopologizing things on characters like fingers and toes and decided that it would work great here too. I brought in a basic cube, then using the make live tool I selected the high poly, subdividing the low poly and making sure it was right on top, then scaling and using the quad draw smoothing tool to loosen the topology.

I used the same approach for the rocks in the chimney.

Texturing

For the wall material of the house, I wanted to be able to generate cracks in the plaster through an ambient occlusion map that I baked out.

The wall base was generated using three different grunge maps with a tile sampler.

I made an input node with a switch so I would be able to plug in my AO map once I was inside Painter.

The AO map was planned to be used to drive where the cracks would show, but I wasn’t happy with how the Ambient Occlusion got baked out so I used an anchored paint layer to drive the mask.

This allowed me to have more control over where the crack would show.

Note: For textures like metal, I did not use any metalness at first. Instead, I found that by adjusting the normals first, you can generate an amazing metallic look.

Final Thoughts

I believe that going into this project with a student mentality set me up for success. My focus has been on character modeling, but by allowing myself to shift my focus, I invited an opportunity to learn about something else. I discovered that I really enjoy sculpting materials!

For the most part, I came into this project with little environment production experience. I do a lot of personal projects on my own time, but working on your own is much different than working with a team no matter how big it is. Learning how to combine your artistic vision with someone else’s and make it look appealing is difficult. Noah and I were pretty lucky because we were almost always on the same page in terms of our visions for the project.

Part II: Noah Kentwortz

Introduction

My name is Noah Kentwortz and I am a third-year student at the Colorado University of Denver in their 3D program. Since I was a kid, I was very interested in how the worlds and effects in my favorite movies worked, and that eventually led me into 3D and compositing. I taught myself Adobe After Effects and Cinema 4D utilizing Youtube tutorials throughout the later part of my teenage years and then transitioned into Maya, then into Nuke after being accepted into the CU Denver 3D program.

Hobbit Hole: Reference

For reference, I scoured both Google Images and Pinterest looking for images. I then compiled everything using PureRef. I highly recommend PureRef to anyone doing creative work, it makes organizing and viewing references easy. It also allowed me to begin building an understanding of the project.

I always have my reference open while I work. Even if I am not directly looking at it, it's sitting at the edge of my vision, feeding my brain information about how the thing should look.

The Fence

One thing I like to do at the beginning of any project is to take an asset through every step of the pipeline to iron out issues that might spring up later.

• The mesh

Once the base mesh was done, I needed to find a quick method of detailing it. Originally, I had thought I would go into ZBrush and detail each piece by hand, but realized this was a good chance to reuse a set of bump and normal maps I had created while detailing the mailbox asset.

• The textures

• Dressing the asset

The Foliage

• Generation

At this point, I had never done foliage and decided to learn SpeedTree. The large number of Youtube tutorials made the program very approachable for a project like this. I used SpeedTree to create a series of grasses, weeds, and flowers, both from scratch and from interpretation and recreation of their included examples which was a really good way to learn the program.

• Shading

In the case of the atlas based foliage, the color map was already there, and it was a simple matter to create bump and roughness maps in Photoshop. The procedural foliage is where things got interesting. For the color map, I used a stretched out image of a leaf and used some nodes in the Hypershade to create variation. I then used noses in the Hypershade to create roughness breakup and bump.

The thing that I feel really made this foliage read well was the subsurface scattering. I was able to use a slightly more saturated version of the color map to drive my subsurface color. As the project was going to be rendered in Arnold, the trick to get the effect to look right on cards was to enable the “thin walled”.

• Layout and Ivy

Although I had originally intended to modify my existing vines setup in SpeedTree for the Ivy, I was introduced to “An Ivy Generator” written by Thomas Luft, which served all my needs on this project. The app is completely free on Luft’s website and I would highly recommend it for some quick foliage or vines. The app contained some basic color and opacity maps for two different leaves. From there, I generated the rest of the maps I needed using Photoshop.

Lighting

My main goal with the lighting was to create a scene that felt fantastical like you could almost feel the air. I was heavily inspired by David Grzesik's Pixar Challenge render of an interior kitchen that fit a lot of the ideas I was aiming for in my lighting. I wanted to make the scene feel inviting and alive and creating that warm interior light with the cracked open door felt like the best solution. I wanted viewers to be able to look at the render and begin to connect with it. Putting the scene at that golden hour, transition time allowed me to push the color contrast in the lights and create the look I was aiming for.

Post Work

Working on a class project meant that, for rendering, speed is even more important than normal. In this case, the main issue I needed to solve in post was the mitigation of noise. I rendered out the scene with light group AOVs, allowing me to isolate the cause of the noise and alter it individually from the other lights.

The main noise source was the emission volume inside the lamp by the door. Where rendering with a higher sample count may have produced a cleaner result, a post fix suited our needs perfectly.

The noise reduction was a very simple and brute force. I isolated the noisy bits by pulling a luma key on them and blurring them. To hide the soft edges this creates, I used a couple more keyer nodes to isolate the sharp, inner core of the lamp and any general high contrast bits and overlayed them back on top of the result. After a little bit of color correction, I was left with a good result.

• The sky replacement

The setup here was simple, I pulled an image that roughly matches the lighting conditions from HDRI Haven (a wonderful site with tons of free HDRIs) and imported it. I then blurred it, did a spherical transform, and then mapped it onto a sphere. I could use the render cam from the original render here, so I imported it and routed everything into a scene, and then a scanline renderer, and finally merged it together with the denoised render. I then added analogue grain and did some color correction and the image was finished.

Final Thoughts

The biggest thing I learned on this project was to render early and render often. Getting that image out quickly and being able to go in and tweak everything really allowed us to push this project.

Block out and use reference for everything. Knowing what I am working on and knowing what I will work on next really allowed me to stay focused and work effectively on things (I have this bad habit of attempting to freestyle a shape completely from scratch, just based on the image in my head, and while it might give me good results occasionally, starting from reference, even if I deviate, kept my work consistent to the style and allowed me to work much more quickly).

Look at things outside. While working, I often get hung up on the scale of things and it slows me down massively. I wasted a good hour trying to decide the scale of tree bark and scratches when I could have just gone outside, found a tree, and took a picture of it while holding my hand up for scale.

Alexius Hill, 3D Generalist/Character Modeler

Noah Kentwortz, 3D Generalist

Interview conducted by Arti Sergeev