Crafting Realistic Dwarven Forge With Blender, Substance 3D & UE5

Introduction

Hi! My name is Louis-Philippe Ross Boivin. I am a 31 years old aspiring 3D Environment Artist from Quebec, and I have just graduated from Think Tank Training Centre. Before doing 3D, I worked as an artisan armorer for almost 13 years at Feral Workshop, which is a company I created with my wife. We worked on creating high-end custom-made costumes and armor for many different events, be it LARP, weddings, historical reconstitution, and theater plays.

I've been an artist pretty much my whole life, and I first started to get interested in 3D around 2019 when I started dabbling around in Blender to create models that I could 3D print for the costume business. After a few months, I realized that I really liked doing 3D and would be interested in furthering my studies in it. With the pandemic giving a really hard blow to most of the events for which our business was creating, I ended up starting a 16-month 3D for games course at College Lasalle in Quebec. After completing that program, I realized that my level of skill wasn't at all high enough to try and apply in the game industry, so in February 2023, I joined the Think Tank Online CG course with a focus on the environment for games.

The Dwarven Forge Project

Ever since I was young, I've always been a big fan of the general dwarven aesthetics that are depicted in many of the great medieval fantasy franchises. Be it Lord of the Rings (which has a gigantic influence on me in both the books and the movies), Warhammer Fantasy, World of Warcraft, and others. So, to close out my journey at Think Tank, I really wanted to create a dwarven-inspired scene.

While looking for concepts and ideas before starting my project, someone close to me gave me the idea of creating a scene that would tie in both my profession and passion, so I started narrowing my field to find some inspiration for a dwarven blacksmith/armor smith shop. I ended up finding this Dwarf Smithy from artist Park Eunji:

While it was my primary inspiration for the general shape of the building and some of the conceptual ideas of the scene (like using water-powered tools), I really wanted to do an interior scene primarily. I continued looking for more ideas, but after spending a few days without finding much inspiration, I ended up looking for some real-life historical inspiration, and it was that part that really tied everything together.

I found a YouTube video from history enthusiast Lindybeige, which showcased the water-powered tilt hammers and other tools that were used at the end of the 18th century in a scythe factory at Abbeydale Industrial Hamlet in Sheffield, UK.

Using those tilt hammers as a potential centerpiece of the scene, I went and gathered some more references to other medieval and Renaissance machinery to inspire my scene.

Blockout & Composition

I started by blocking out most of the exterior of the building by following the concept I had chosen. At this point in the project, I was still planning on doing both the interior and exterior of the scene, so most of my focus was on getting the general shape of the building right.

Since I hadn't found a clear reference for what I wanted to do with the interior, I used the furnace placement that was visible in Park Eunji's concept and built the rest of the interior based on that placement and the position of the waterwheel.

At first, the general scale of most of the big tools was quite a bit smaller than what I ended up with at the end. The scale I had in mind at first was a little bit smaller than most of the historical equivalent since I was building a dwarven building, but after a few back and forths with my mentor throughout the modeling process, I ended up scaling up many of the mechanical pieces in the scene.

I was a bit ambivalent about doing that at first, mainly because I felt that those smithing tools needed to be smaller to be efficiently used (in hindsight, it's pretty funny because I was still in a mindset of trying to create a somewhat realistic smith workshop) but after a few tests, I realized that the scene gained a lot by having those assets a lot bigger than what I had planned at first.

It did cause a few problems down the line; with the new scale, I had to modify the general layout of the scene and make the building bigger (while trying to work with the modular assets I had already created for the building). This process was repeated a few times until I realized that with all the scale and layout modifications, the scene had lost some of its initial coherence and was now a bit all over the place. When I realized that I was about 3 months into the project with many assets already completed, after some discussion with my mentor, I decided to redo most of the layout to focus the scene on 3 specific sections. At that point, I made the choice of reducing the amount of water-powered hammers to one and centering the furnace to make it a focal point of the scene.

Doing that, I had much more space to work with on the left side of the scene and chose to add some more forging and metalworking tools that I hadn't planned at first. In that layout change process, I also created a new asset kit for the ground of the forge and used it to create the small stairs on each side of the furnace section.

Modeling & Topology

From the start, one of my goals for this project was to create as many complete assets as possible to practice the general pipeline of low-high bake with clean UVs and texture them in Substance 3D Painter. Except for two or three parts of the scene, all the other assets used some sort of high-to-low bake.

My modeling workflow is always pretty similar, I start with a general blockout of the piece and then apply some basic materials just to get a basic idea of the look of the piece. With those materials on the model, I also do some basic UV, mainly to be able to visualize at least part of the material while I am modeling. At that point, I will probably start working with a simple bevel modifier and a weighted normal modifier just to quickly get an idea of the general look of the piece.

Using those, I will go on to model the more complex part of the piece. Most of the assets in the scene were created using this method. Unless I know that I will need to do some complex sculpting or some heavy damage, I usually create a mostly complete low-poly first.

Unless it is unavoidable, I rarely completely retopologize an asset by hand. Sometimes, I can't avoid it, but using some good low-poly methods and the Shrinkwrap modifier in Blender, I am often able to reach a result that is more than good enough for baking. Sometimes, some hand adjustment is required on a model after using the Shrinkwrap modifier, but it is quite useful to quickly get a solid base to work on.

For UV unwrapping, I do most of the work in Blender using some great add-ons, mainly ZenUV, UVPackmaster, TexTools, and Texel-Density. I also used Rizom UV on this piece to use its powerful auto UV capability to unwrap the coal pieces I used in both the furnace and the box.

In general, I start UVing as I model the piece. I learned to UV by hand marking my seams by hand, so I rarely use auto UV and projection UV because I prefer to have more control over where my seams will be. The main tool I use for most of my UV manipulation, like straightening and quadrifying, is, without a doubt, ZenUV. If I had a UV Blender add-on that I would recommend the most, it would be that one.

Texturing

To keep in line with my goal of practicing the standard asset pipeline as much as possible, most of my assets in the scene are textured as hero props in Substance 3D Painter with a 2K or 4K texture. While clearly not optimal if it was a scene that needed to be rendered in a game, it was perfect for my goal.

For texturing in 3D Painter, my biggest gain during this project was studying and understanding how Javad Rajabzade built his many materials and smart mask. Understanding how to use multiple small and thin layers of detail on a full prop to reach a great final result.

With the release of Unreal Engine 5.4 and the new material integrated displacement function, I really wanted to make use of it in my scene, so I did some trial and error to get it functional with my texture's height map. It was a bit of a process because the Displacement calculation from the UE shader isn't the same as the height displacement in Substance 3D Painter, so I had to use some nodes to adjust it and get it to work.

With this, I was able to get some nice lighting and shadow on my flat wall and on some of the brick parts of the furnace. Working in conjunction with the normal map allowed me to get some nice depth on those sections of the scene.

In addition to that, since I was trying to create a grimy blacksmith shop, I wanted a way to add grime and dirt in the engine so that I would be able to change it with more flexibility than having to go back into Substance 3D Painter each time. Since UE5 Nanite doesn't really allow vertex painting (there's a workaround, but it's not really optimal), I tried to find a different solution to do what I was trying to do. After talking with a few of my peers at Think Tank, one of them suggested to me that I look into distance field ambient occlusion and that it could maybe help me. And he was right.

Using the DistanceFieldApproxAO node in the Unreal Material, I was able to create some dark grime and multiply it into my base color channel. It was perfect for my needs and easy to customize with some basic parameters.

Composition

When doing the final composition for the renders, I still tried to keep in mind that I wanted the scene to be something that could be realistically used by a dwarven smith. It was a bit tricky at some point because a real workshop can be really messy and unreasonable, and it wouldn't have been great for the shots but I still tried to place things in a way that would make sense.

All assets were placed by hand and rearranged a few times for certain shots. It was always a challenge of trying to get as much of the created assets in a shot as possible without losing the focus.

Lighting & Rendering

Lighting, in general, has always been a big weakness of mine, so I received tons of help from my mentor on that part. I had a lot of different iterations of lighting for the scene, and I would say that most of them were rough.

At the beginning of the project, when I was planning on doing the exterior of the forge, I tried using Ultra Dynamic Sky, which is an Unreal Engine toolset to create dynamic lighting with sun and moons but also snow and rain, which I wanted to use. While a lot of the basic settings of UDS are quite similar to the basic lighting settings of a normal scene, there were some quirks that weren't really appropriate and were causing more problems for me than what I was gaining from it. So, after choosing to do only the interior of the forge, I stopped using UDS and rebuilt all of the lighting in the scene using the standard Unreal Engine 5 toolkit.

One of the great tricks my mentor showed me was to turn up the Shadow Source Angle Factor by a lot in the directional light to greatly soften the shadows created. With Raytraced shadow on, I had some difficulty adjusting the shadows to get them softer and changing that setting alone was a great help in fixing that.

Since I wanted to use Lumen and raytraced shadow in the project, I also had to deal with the qualms coming with it. One of those qualms is that raytraced shadow on basic settings kills most god rays and light shafts by softening most of it.

The solution to that problem was to use some rectangle light in the windows where there would normally be god rays and make them each have a really high Volumetric light dispersion with a really long light length. With that setup in place, it was possible to manually adjust each god ray coming from each window individually without having to change the big directional light of the scene each time.

Those setups also worked well with the atmospheric particles that I added to the scene to simulate some of the floating dust using Niagara. By tweaking the standard particle materials and using basic color instead of emissive, I was able to make most of the floating dust only appear inside the direct light rays coming from the windows and from the candlelight.

For the furnace, I added two particle emitters to simulate the floating embers that would be coming from the burning coals and one for the smoke. A large part of the final lighting from the scene came from the HDRI backdrop, which I added to bring some soft blueish lighting from the outside. Since the start of the project, I have always had in mind that my forge would be in some sort of boreal pine forest in late winter/start of spring, so I downloaded an HDRI representing that kind of environment. It really helped bring the whole scene together a lot.

Conclusion

The main challenge overall was probably scope. Since I was working without a fully established concept and creating as I was going, it was hard to keep a good handle on the amount of work needed to bring the scene to completion. The multiple change of layout of the scene also blurred the lines that I had established at the start of the project, making time planning and management way harder than it should have been.

The second challenge and lessons learned would probably be to do more lighting and rendering earlier in the project. I ended up doing a lot of work that I had to go back and fix or even redo in some cases because the lighting was clashing quite a lot with some of the textures I had made. This was in part caused by my decision to abandon the use of UDS a few months into the project, but overall, I could have spent more time doing some serious lighting test before creating most of the final textures like I did.

Advice to give would really be to get external feedback while working on a scene. Looking at a concept/work every day for weeks can skew our perspective a lot. Getting some fresh insight from other people can sometimes help a lot. Getting a mentor is, of course, probably the most efficient way to receive serious 3D feedback, but even just getting the first impression feeling of people around you looking at the scene may bring some out-of-the-box insight and help put in perspective some mistakes or problems with the scene that are sometimes invisible to you after looking at it for so long.

Overall, this project was a great opportunity for me to learn and practice many techniques. The fun part of 3D for me is that there's pretty much no end. The more I learn about it, the more I realize that there are many more things that I don’t know about or don't really understand, and it's a really great motivator for me to continue working and digging.

I want to thank Forrest Mead, my mentor, for all his great help during this project and many other Think Tank students who helped me and gave me feedback throughout the whole creation process.

Louis-Philippe Ross Boivin, 3D Environment Artist

Interview conducted by Gloria Levine