3D Modelling in VR
In the past ten years, the workflow for creating 3D assets has fundamentally changed. Previously, modeling and texturing were two separate processes: first, the 3D model was built and optimized, then the UV map was created—and finally, the completed model was textured. The focus was always on the polygon count and texture resolution. However, the increasing importance of sculpting programs like ZBrush, along with the use of low-poly and high-poly models, has drastically changed the requirements for developers and artists, completely overturning the traditional workflow.
3D Modelling Workflow
In the past, the 3D modelling workflow was quite straightforward, with individual steps clearly separated. Once the modelling process was completed and the model approved, the artist had to create a UV map, which is saved as a 2D texture and acts as a bridge between the 3D model and the texture. UV maps needed to be easily readable, with adjacent parts of the model also placed next to each other on the texture. Additionally, it was important for all UV shells to be uniformly scaled to avoid significant differences in texture resolution.
Following that, texturing began in a traditional image editing program like Photoshop. If substantial changes were required for the model after texturing, the corresponding UV map had to be adjusted, and the necessary changes had to be transferred to the texture. Therefore, it was already crucial to align the model early on with clients or team members to avoid unnecessary modifications. In this classic method, there are still only two main processes—modelling and texturing—followed by the final UV mapping.
High Poly and Low Poly
The modern high poly and low poly workflow is significantly more complex, with the steps being much more interconnected. Each individual step must be executed carefully and diligently, as mistakes inevitably affect all subsequent processes and are often difficult to correct without starting over from the beginning of the workflow.
We start with the high poly model, which we create in all its details and with relatively few technical limitations in a sculpting program, or—in the case of hard surface objects—in a traditional 3D modelling program such as 3D Studio Max, Maya, or Blender. Once the high poly model is approved, we can begin working on the low poly model. For this, we use the high poly model as a reference and build the low poly model as an outer shell. At this stage, we must consider the requirements of the real-time engine and keep an eye on the polygon count.
Furthermore, it is crucial that the overall shape of both models closely matches. The low poly model should penetrate the high poly model as little as possible. Only in this way can we ensure that the texture is transferred during the baking process without artifacts or errors.
... to UV Mapping
The next step is UV mapping, which we are already familiar with from the classic workflow. However, there are some key differences here. For example, it's important to cut the UVs at all 90-degree edges to create separate UV shells. Otherwise, black seams will become visible during baking, overlaying the texture. This can make the UV map quite intricate and confusing, especially for inorganic models with many edges. Typically, this would create significant problems, and such a UV map would be impossible to use for texturing in Photoshop. Fortunately, modern 3D texturing tools like Substance Painter can read and edit auto-generated and completely complex UV maps. Additionally, once the UV map is created, it's essentially finished.
This actually simplifies the UV mapping process. Thanks to various auto-UV tools, we now only need to place seams on the hard edges and no longer have to worry about further dividing the UV shells. However, a different process comes to the forefront: creating a cage, which we will need later for baking. During baking, we transfer the high-resolution details from our previously created high and low poly meshes onto the performant low poly geometry by emitting rays. For texturing and baking, we use Substance Painter. The advantage here is that the textures are correctly named and automatically organized in Substance Painter. After baking, we can proceed directly with texturing.
We place the cage model around the high poly model like a protective outer shell. The cage must have the exact same topology as the low poly model. In most cases, it suffices to inflate the low poly model with a push modifier. This helps avoid artifacts and errors on the texture.
When the high poly, low poly, and cage models are correctly set up, we let Substance Painter bake the textures. We can immediately inspect whether everything worked out in the 3D viewport afterwards.
3D Texturing with Substance Painter
As mentioned, Substance Painter is a 3D texturing tool that allows us to edit the 3D model directly with various brushes, offering a layer hierarchy similar to Photoshop. However, Substance Painter's particular strength lies in its procedural textures. During the baking process, we have access to different textures, such as normal, ambient occlusion, and curvature maps. Using these maps, we can mask out elevations or depressions while texturing, meaning we can focus on areas that are above, below, in a specific cardinal direction, or at a previously defined angle.
For example, we can select all edges and highlight them with scratches, apply rust to lower areas, or allow moss to grow on a tree in just one direction. Substance Painter provides us with many layers that function like layer compositions and can be applied to any model. It's also possible to convert our custom layers into a Smart Material layer group to share with other artists. After texturing, our model can also be lit, and we can bake lights into the texture. This approach is especially essential for mobile applications with few or no lights.
In short, Substance Painter is very complex but offers the user many possibilities. With a bit of practice, it's possible to create textures of a quality that would be difficult to achieve in Photoshop.
Working in ZBrush
When combined with Substance Painter, ZBrush produces the best results. What sets ZBrush apart is the high polygon density of the models: in conventional 3D programs or real-time engines, artists must always pay attention to the polycount to avoid jeopardizing performance. In ZBrush, we can easily bring the polycount to several million per model. Practically speaking, this means we can effortlessly create a character made up of 50 individual models, each with a polycount of 2 million, without significantly impacting performance.
Of course, we can't manipulate or move this vast number of vertices individually as we would in 3D Studio Max. Instead, we use brushes and tools that are reminiscent of traditional sculpting tools.
Working in ZBrush feels very direct and intuitive: we work with a kind of digital modeling clay, allowing for detailed work without having to worry much about performance or technical restrictions during sculpting or high poly modeling. Only when the high poly model is completely finished and approved by the client or art director do we create a low poly model and transfer the details onto a texture. This altered approach has had a significant impact on our work as 3D artists, as it has suddenly opened up many new possibilities and challenges.
Even education changed as a result: artistic disciplines like drawing, sculpting, and manual modeling have suddenly become very important again and have been integrated into the curricula of many training institutions, universities, and art schools. Nonetheless, technical understanding remains essential, and without a clean low poly mesh, we ultimately won't achieve a good result.
“The high poly and low poly workflow has become increasingly important in recent years and is essential in modern 3D productions. Although it is now much more complex and, therefore, more prone to errors than classical methods, the advantages clearly outweigh the disadvantages in the end. Textures created from a high poly model look extremely high-quality even with minimal effort in texturing and can be easily optimized. The individual steps are clearly separated and can be easily distributed among different artists or outsourced.”
Sebastian Wunnicke, Senior 3D Artist
Would you like to learn more about 3D modeling?
