Even though we plan to continuously expand the variety of supported 3D content, Umbra by nature performs better in some circumstances than others. Some of the relevant characteristics are listed below.
Umbra performs well on scenes with the following characteristics:
- Large triangle counts with redundant data viewed from long distances.
- Unique objects and textures.
- Heavy occlusion.
- Solid objects.
Scenes with these characteristics often get an average performance increase of around an order of magnitude.
Umbra performs poorly on scenes with the following characteristics:
- Low triangle counts or low complexity in general (plane that consists of two large triangles):
For scenes with low complexity, it is possible that the results that Umbra generates will have more complexity than the original scene.
- Objects with very high frequency geometry and holes (fences, tree branches, grid mesh):
High frequency geometries might lose detail after the computation is done, especially if the smallest feature parameter is larger than the small details.
Umbra currently has some restrictions on input scene contents that prevent computation:
- Before computation begins, the scene has to be imported into Unity. Therefore the scene has to at least open inside Unity, which itself can cause issues with certain models.
- Extremely complex regions in the scene and high resolution textures might cause the computation to fail due to memory requirements.