Why this?
Technical Artists come from different backgrounds, but the two main ones are usually computer science and art. Both excel in different parts of game production. While artist-background TAs may not always go as deep into code as their more technical counterparts, in my years working in AAA I have seen the same problem again and again: Technical Artists who overload their tools with options, creating monsters that artists cannot really control. At that point, the tool stops being a tool and becomes a constant stream of calls to the TA team to “fix” it.
I think the foundation of a good landscape material should be built with the artist in mind. It should be easy to use and extremely simple. Every option in the material should be something the average artist will use again and again, and it should be understandable immediately. The rest of the fancy and cool effects should be extras, added later if the artist requires them.
That was the mentality behind this material when I started working on it. I wanted to create the skeleton of a landscape material that a AAA studio could realistically use in an open-world pipeline. The material is modular and easy to expand. It currently uses two biomes and three layers per biome, but adding more layers or biomes should be a relatively simple task. More importantly, the artist should know exactly what each option is for. The material also uses a mask system to assign Physical Materials to the different layers, with the idea that they will be used later by the PCG system.
High-Level Architecture
One of the most interesting aspects of this material is the relationship between layers and biomes. The setup uses 3 layers and 2 biomes, but instead of giving each biome its own separate paint stack, the same three layers are shared across both. The workflow is simple: a base layer covers the entire landscape, and two additional variation layers are painted on top of it. A biome mask then decides whether each area resolves as Biome 1 or Biome 2.
This approach may look a bit strange in a setup with only 3 layers and 2 biomes, but this material is meant to be the base for a much larger system. Once you imagine 5 biomes with 5 layers each, this kind of structure makes a lot more sense.
The main advantage is that, by restricting how much artists can freely mix biomes and layers, the material becomes more robust and works better with the PCG system. Biomes are meant to define large areas, while the layers act as variations inside them. It is also much less intimidating for the artist, since they only have to choose between 3 to 5 layers instead of dealing with 25 different possibilities.
The biomes themselves can either be painted as a normal landscape layer or driven by a mask.
In the image above, we can see how the material is structured. Both Biome 1 (Desert) and Biome 2 (Forest) are processed first by blending the base layer with Layer 2 and Layer 3. Once both biome paths are resolved, they are blended together using the biome layer or biome mask. After that, the material applies the optional cliff layer based on slope, followed by the optional snow layer based on both slope and height. The final step in the main chain is the macro variation pass.
There are also two sections outside the main chain. One is the RVT output, which records the final material into a Runtime Virtual Texture. The other is the mask generation section, which creates all the masks needed for the Landscape Physical Material Output. This structure keeps the material easy to expand and ensures that the same terrain logic drives the visual result, the Physical Material classification, and the RVT output.