- MEDIA ˇ
- GAME ˇ
- DEVELOPMENT ˇ
I’m currently experimenting with an automated realtime LOD (Level of Detail) generation. Normally in games the level builders will create a simpler version of the objects used in the game to be shown when far away. This is needed to keep a high framerate as the number of objects that are visible increases with the viewing distance, and the number of objects and their complexity directly impacts the performance of the rendering.
As we plan to have destruction and fires built into the game, we require a system that can capture the result of these modifications to the world instead of having the level builders doing a predefined set of objects, hence the automated realtime LOD system that I’m currently working on.
Here are some very early first screenshots showing a technique called volume slicing where the model is projected onto a set of billboards that is then drawn instead of the objects themselves. The nice thing about this technique is that it doesn’t require any expensive analysis of the objects. The bad thing is the cracks that is obvious when viewing the model from angles that are not orthogonal onto the billboards. I’m currently considering to combine this with relief mapping to reduce the cracking.
Here are a few links to papers that deals with automated LOD generation
After Henning implemented the vector-to-scalars and the Vector_State->Intersection_Point this stuff is possible. The shader blends some green mushy algea stuff on the lower part of the object so the effect will always be in line with the water on the level no-matter where the the object is placed :-)