Open-world games are immersive and expansive, but they also pose many challenges for 3D graphics developers. How can you create realistic and diverse environments without compromising performance and quality? In this article, you will learn some of the best practices for optimizing 3D assets in open-world games, such as:
LOD is a technique that reduces the complexity and size of 3D models based on their distance from the camera. This way, you can save memory and processing power by rendering only the details that are visible to the player. You can use tools like Blender or Maya to create different LOD versions of your assets, or use automatic LOD generation in game engines like Unity or Unreal.
It’s important to make sure you setup the correct distance between LOD version to ensure a smooth transition, this will help to avoid breaking the inmersión of the player. The LODS can be powerful and tools on Maya or Blender offer options to do automatic retopology and reduction of geometry. You can use Reduce on the newest version of maya and have a clean reduction avoiding to distort too much the Silhouette Silhouette might be one really important, avoid doing drastic changes of it during the LOD creations. Also during production discuss the time that should be invested on this LODS, are a fundamental part, and it’s important to make them for optimization.
Unreal | Unity | Technical Artist | 3D Generalist | VFX | AR | VR | Archviz | Tour360
The most common approach in Engine could be to implement Level of Detail(LOD) in combination with other techniques like occlusion culling/pre-compute visibility, mesh to Impostors etc Other approaches in 3D software packages could be decimating the meshes, deleting non-visible faces, and avoiding overlapping of their hidden faces. In terms of performance and optimization, the most important is planning and implementing a good workflow based on a budget and target platform. Usually if is left this for the end could create bottlenecks during the production, and probably retouching or even redoing assets. The optimization of 3D assets is not only about poly-count but also means creating good UVs, Textures, Materials etc. Everything count
On top of maintaining a consistent silhouette through out your LODs make sure the main features of your model are also distinguishable from a distance.
I prefer to use only automatic LOD, and very rarely I model more then 1 high-poly and 1 low-poly models. LODs can simply made by another person as Tech-artist or engine-developer So, making different LODs by myself usually not my way for optimization
Use Imposters: Replace complex 3D models with 2D billboard sprites or simple 3D shapes when they are at extreme distances. Asset Streaming and Management: Implement an asset streaming system to load and unload assets dynamically based on the player's location in the world. Prioritize the loading of assets that are essential to gameplay. Optimize Mesh Geometry: Remove unseen polygons, such as the bottoms of furniture or the backs of buildings. Keep the geometry as simple as possible while maintaining visual fidelity. Consolidate and Share Materials: Minimize the number of unique materials and shaders used across models to improve batching and reduce the number of draw calls.
Textures are one of the most memory-intensive components of 3D graphics. To optimize them, you can use texture compression formats like DXT or ETC that reduce the file size and bandwidth without losing too much quality. You can also use texture atlasing, which is a technique that combines multiple textures into one large image, reducing the number of draw calls and texture switches.
Well the atlas is one of the best alternatives, seeing the most optimized packages on unity is clear that they are using whole worlds painted with a single atlas, using colors taken on each UV to make sure they can paint everything super fast. Targa might be better in case you have textures that you really want them with the whole information, as well as using shaders to combine different maps on a single channel. And it’s true what it says here, avoid making multiple calls of textures, reduce the amount of the materials that are being called, if you are doing hero props. You can pack lots of information on a single texture, but keep on mind the texel density to avoid having weird sizes or quality issues between maps.
Choosing the right type of texture format will help you decrease your memory and storage footprint. Ask yourself questions such as, do I need an alpha channel or is 8bit color depth per channel enough or do I need 16bit?
In my experience, there a lot ways to optimize with your textures. - You can draw high resolution maps, and then, if it is necessary for project, you can compress all Maps - It can be the strategic of modeling and texturing from begining, as trim sheet / color map / or atlas for group of model - also you can separate alphas or main details to another UV-map with different level of tessellation for it I like this type of technique, and using a lot in my work
When it comes to texturing it's really important to choose correct size for our textures. Depend on the texel density of model we have to choose our texture size.
I found it useful to reducing resolution for non-crucial details, optimize color channels, testing out JPEG compression for non-realtime applications, removing unnecessary fine details, reducing colors, simplifying or partially removing unnecessary alpha details, and exploring general lossy compression algorithms. Always test your texture to strike the right balance between compression and visual quality so as not to make it too compressed or too complex
Occlusion culling and frustum culling are techniques that prevent rendering objects that are not visible to the player, either because they are behind other objects or outside the camera view. This can significantly improve performance and reduce overdraw. You can use spatial partitioning methods like octrees or quadtrees to organize your scene and determine which objects are occluded or culled.
Batching and instancing are techniques that reduce the number of draw calls by grouping or reusing similar objects. Batching combines multiple objects that share the same material and mesh into one draw call, while instancing renders multiple copies of the same object with different transformations and parameters. You can use static batching for objects that do not move, and dynamic batching or instancing for objects that do.
For artists, my biggest tip is to avoid using multiple materials per object unless it's a really large object that requires separate materials. Another technique that some people use when creating prefabs is merging them into a single mesh. This way, your prefab includes all the meshes you need, allowing you to avoid placing each mesh individually but still keeping them separate in groups.
Here are some tips, you can: 1. combine similar materials and simplify meshes. 2. use texture atlases to minimize materials. 3. Implement LOD and frustum culling. 4. try GPU instancing for better rendering. 5. merge similar objects and compress data. 6. explore geometry shaders and multithreading for efficiency.
This way is very useable as in 3d-program, as in game engine. If game logic need a lot of requestes, it can make worse problems for optimization, as lot of polygons/tris, texture and fail memory store or physics and simulation. So, it is 3d-artist resposibility: - organize and group objects into one mesh (when you can), produce readable naming, - make more simplify assets and prefabs, - create less keys in animations, - delete all unnecessary objects, info, mesh, materials etc. - and last you can add custom joints to objects or make blanshape for easiest and fast way for manipulating with your models in game-engine
Lighting and shadows are essential for creating realistic and atmospheric scenes, but they also require a lot of computation and resources. To optimize them, you can use baked lighting and shadows for static objects, which pre-calculates the light and shadow maps and stores them in textures. You can also use dynamic lighting and shadows for moving objects, but limit the number and size of light sources and shadow casters, and use low-resolution or cascaded shadow maps.
Make sure to put all your objects that are not going to move on static before baking. Making your light maps resulted from your baking are correct, sometimes light maps can indicate problems with too much exposure or too little. Also, keep baking every time you can, don’t want till the end to bake, since this can create lots of problems that you are not calculating, and it’s one of the worst experience you can have when baking for the first time. Be careful with double shadows when baking, make sure your light makes sense, this can break player’s experience.
Lighting is one of the most math heavy processes in the 3D world. There are constantly improvements that help reduce this strain on the resources. Be sure to test over and again your lighting systems until they look good and are super low on resources requirements. Lighting is the most important part of creating immersion in a 3D scene.
When it comes to lighting and shadows, always prioritize real-time shadows where they have the most impact, optimize shadow quality based on object distance for efficiency. You can also consider temporal anti-aliasing for smoother dynamic shadows, use geometric proxies to streamline dynamic lights, enhance directional light shadows with cascaded maps, apply screen space reflections selectively for efficiency.
LOD bias and mipmaps are techniques that adjust the resolution of textures based on their distance from the camera. Mipmaps are smaller versions of textures that are used for far-away objects, reducing aliasing and moiré effects. LOD bias is a parameter that controls how much the texture resolution is reduced, depending on the screen size and quality settings. You can use LOD bias and mipmaps to balance between performance and visual fidelity.
This is really powerful indeed, the combination of these two can help you get really good performance, do planning of all of this on the way and don’t leave this till the end, lots of projects suffer from this part that is critical on the art side for optimization, as important as the lighting baking or code optimization.
Always carefully consider the placement of origin points. If you are making an exact replica of an environment, place the origin points of each static object at the same position. So they can simply be set to 0 and everything matches exactly. If you are building something modular, think about what would be the best place to put an origin point, so that objects can be easily instantiated in a logical way. If you have interactable objects, depending on their function, either place their origin points on the center of rotation or in the bottom center. And don't be afraid to add empties as rotation axes. Each of these will save your teammates (or yourself) a lot of headache when implementing your models into an engine or other 3D projects.
One thing I found helpful for me (and this is more about animation systems) It is very comfortable for both developer and artist, if you add to your tech-model (engines, cars, gun reloaded system etc.) joints into 3D-editor, and also blendshape for simply animations or just a little variations for, as example, face animation You or other person can with this tools make fast animations for your projects. In some situations, this a most fastest way, in contrast to classic rig or solid animation. However it requiers soft skills, and a bit more time in game-engine