The Legend of Hyrule’s Castle

Introduction

This was my first Midterm here at PIXL VISN Media Arts Academy where I had to make an art piece out of a given theme, and also the first time I was allowed to focus on a more specific part of the production pipeline. Mid/End Terms are intense 1 week training projects.

The given theme this time was “Nintendo”. This threw me out of my concept at the beginning since I never consumed a lot of their content. After researching The Legend of Zelda a bit more (which I had never played before… I know what a deadly sin), I finally decided on creating a realistic version of Hyrule’s Castle after an amazing concept art by Raphael Lacoste. Without further blabbering I present to you my version of Hyrule’s Castle:

Academy Project Guideline

We had four different projects to choose from, with each of them focusing on different aspects of the Pipeline. My choice was “Environment with built structure” in which I had to create an environment which is either full CG or a mix of CG and compositing. I was allowed to download and use all kinds of assets, leaving only the built structure which had to be modeled and textured by myself. After one week I then had to deliver at least 5 seconds of rendered footage excluding titles and breakdowns.

Software Used:

-  Maya → hard surface modeling
-  Redshift for Maya → turntables
-  Substance Designer → trim sheets
-  Substance Painter and Designer → banner texturing
-  Houdini → banner simulation
-  Unreal Engine 5.3 → scene assembly and rendering
-  Photoshop and Davinci Resolve → video / photo editing

Personal Goals for this Project:

-  Focusing on only 1 shot and making it as good as I can.
-  Creating a fully functional modular set using as few materials as possible.
-  Nailing the composition and layout of the shot.
-  Getting into the workflow of Speed Tree and Unreal Engine for Vegetation.

References and Planning

From my own experience and every teacher / tutorial telling me a thousand times it is pretty clear: REFERENCES ARE EVERYTHING!

So, I went ahead and collected all the reference material I thought was necessary, creating a mood board, and analyzing the concept art from Raphael Lacoste to understand the composition and his idea behind it. When looking for references I usually browse websites like Artstation and Google to find what I'm looking for. I recently started using Miro instead of PureRef to collect all my references and I've been really happy with it so far. The benefits I see here is that it also allows me to add videos to my collection and even share the board with instructors or other students to work together.

The last step of my preparation is always to create a Time-Management plan so that I know what to do roughly on each day:

Blockout

A very important difference between environments for Film and Games is that in Film you almost always have a fixed camera angle. This seems very obvious but when taken into consideration early can save the artist a lot of time. (This is because you don’t have to build anything that the viewer won’t see, and all the detail can be managed according to the distance between object and camera.) For that reason I added (to the render) invisible barriers into the scene that would stretch across the whole world. It told me at any point of the project what was visible in the final shot.(Huge time saver!!!).

With that in mind I started to work on my first goal which was to create an immersive composition for the shot. Working from concept art a lot of that was already given, which meant that most of the effort went into the camera movement and playing around with scale. Since the Mid Ground in this shot was basically just a large valley it was very important to create some depth between foreground and background. This turned out to be one of the major issues with this project and needed a lot of adjusting. The key components that worked in the end were: a camera movement that had a lot of parallax (foreground and background seemingly moving at different speeds for the viewer), extending the foreground to give it more depth and finally adding different moving elements like smoke and birds to have something that indicates the valley.

After placing a few cubes and a little foreground terrain this was the first version:

Hyrule’s Castle - Asset creation

Modeling

As I was aiming to create a fully modular castle set, (which should also save me a lot of time modeling and texturing a huge amount of assets), I first started by color coding the original Reference to get an idea of which pieces I had to model, and how they should fit together.

After that I brought my basic shapes from Unreal Engine into Maya and rushed into the modeling process which in of itself went quite well. Knowing that this asset will be in the background I held back quite a bit on the details I put into it and tried to keep the poly count rather low.

Texturing

When it was finally time for UV’s and texturing, I had to experience my lack of knowledge in this kind of field. The goal was to use as few textures as possible, and not just go with the workflow of texturing all the modules separately as I had done multiple times before. I wanted to create 2 trim sheet textures for the Castle and use them for all the modules I had created. However, I learned over the course of this project that texturing and modeling of any asset should go hand in hand to achieve optimal results, especially when working with trim sheets.

So after this realization I started creating trim sheets in Substance Designer for the modules I had made in Maya. This was a back and forth process of creating a repeating pattern in Designer, exporting the maps to Maya and laying out the UV’s according to those maps. For this testing process I used basic noises and generators to make sure all the tiling on the modules will be working.

For the brickwalls of the castle I faced the issue that the usual horizontal tiling of a trim sheet was not big enough to actually get the brick size and detail I was looking for. This is why I decided to use a node called “Non-Square Transform” which allowed me to combine different non square textures that were tiling in u an v into one square texture while maintaining the tiling of each patch in u an v. (Tutorial for this technique: https://www.youtube.com/watch?v=l6l_6V7kpeA)

Looking back on this decision I do think it was kind of stupid. For some parts of the asset I should have just used regular square tiling textures, because there was no benefit of fitting them into a trim sheet if they should be tiling in both u and v anyway. (It was a really great and intense learning experience though.)

After the technical part of assigning all the different trim parts from the textures to my model was done, I continued to actually create all the different materials in Substance Designer.

Scene Assembly

Going back to Unreal Engine, let's have a look at the world building part of my project.

So after I had the base model for the castle done I did the first test of importing the asset into Unreal Engine and creating a Layout to see if it was all working together. That is when I decided on a few last tweaks in the modeling before moving on to the texturing part just described

The next step then was to work on the foreground terrain. For the Blockout I had used the Unreal Landscape tools and sculpted it to the rough shape I was looking for. I now used different ground assets and plants from the Quixel Megascans Library to populate the foreground landscape. Also I decided to add this big boulder rock formation in the foreground pointing right at the castle to enhance the composition by giving it sort of a guiding line towards the castle. 

After that I added the final castle modules to my Unreal project and matched the layout I already had in the scene. 

This was when I was really starting to have a lot of fun with this project. Next I added the cliffs in the background using Quixel Assets, and also worked quite a bit on the lighting and atmosphere in the scene. I played around with the sky and cloud movement as well as added some local height fog to increase the foggy fade coming up from the valley. (https://www.unrealengine.com/marketplace/en-US/product/ultra-dynamic-sky) 

I really love these round full bushes and trees that were growing all over the cliffs in the reference. My original plan was to use Speedtree to custom create them and match them as closely as possible. Sadly due to time constraints I had to scrap this plan and decided to get a really nice, kind of matching, bush kit from the Epic Marketplace. I then used those to scatter them all round the cliffs in the background.

The last steps were to really enhance the shot by giving it a lot of movement and life. For this I used different elements: Birds flying in the sky and up from the valley, animated fog cards in the valley and also inside the castle and finally the two foreground banners waving in the wind. Also I added a lot of decals and some color correction (CCR in Unreal) to the castle to enhance its textures because it was feeling (and partially still is) really flat. The last big change I made here, (which i had already teased in the beginning,) was to extend the foreground a lot which gave the shot a lot more depth. 

One more compositional thought I’d like to share is the placement of the banners in the main shot. Originally there was only the banner on the right. I then decided to add another introduction shot that would show the banner in the left third of the screen space. However the cut between the two shots then caught the eye of one of my instructors who taught me the importance of compositional continuation between shots. This meant adding another banner in the foreground roughly where the banner in the first shot was. This kept the viewer’s focus in the same spot. Then use the composition of the second shot with the banners, arrows and rocks etc. to guide the viewer's attention to the castle.

Rendering, Compositing and Video Editing

When it came to rendering I kept it rather simple. I rendered just a basic beauty pass, as well as Z-Depth and Cryptomattes (Object ID). I then brought the shots into Nuke and did some very basic compositing on them, before exporting them into Davinci Resolve where I did the final video editing.

Link to the Final Result:

https://vimeo.com/914219801?share=copy

Conclusion

This was a really intense but also beautiful creative week. In the beginning I really hated the outcome and all its flaws. (and believe me if you can’t see them right away, there are quite a few) But over time I started to appreciate how it turned out and especially all the things this project has taught me.

In addition to me pushing through the challenges, feedback was a very essential part of this project. There are a lot of creative and technical decisions I couldn't or wouldn’t have made without the help of all my colleagues and instructors along the way. So I want to use this opportunity to thank all of you for your time and support, especially Fabrizio Meli, Robb Innes and Maksym Osmolovskyi.

Thanks a lot for taking the time and reading this post, and please feel free to reach out anytime.

Raphael Stiep