I’ve viewed nice tuts on how Volume Streaming Levels function, am prepared to organize my Maps accordingly, but I’m left to question what workflow on the front end in developing environment assets both maximizes quality and optimizes system resources.
I’m working with image scan data to produce highly realistic virtual environments, the emphasis being on realism of the experience, no slinging swords and involved game play. Interactive content comes second, system resources are mainly about image fidelity. My source imagery in 3D capture is 52 MP, use photogrammetry to model a complex cave system, SFM software allows me to export large sections of cave (limited only by system resources to reconstruct) with control over breaking a monster mesh into X number of smaller nested meshes based on max vertices per part. These nested sectional meshes share a common scale and coordinate system, dropping into UE4 seamlessly. I then also have control over the size and number of texture maps used to cover the larger “Reconstruction Zone”. My thought is to kick out X number of 8K textures (the limit in UE4, my deliverable assumes beefy system resources), but to break the geometry into relatively tiny mesh parts to leverage occlusion culling, LOD and such to optimize system resources what gets drawn to the screen by GPU. I’m eager for feedback on the validity of this strategy.
Reconstructions Zones are scaled and positioned numerically same as you position Maps within Level Streaming Volumes, so I get the outlines of how I’ll want to pair adjoining Reconstruction Zones with adjoining Maps within overlapping Level Streaming Volumes based on size and position. Understanding that LSVs are leveraged to optimize system resources, I’m left to define what it means to max image fidelity (with due respect to the complexity therein) against optimization in generating assets. How would you go about determining how many mesh parts a given volume is best broken into, how to set max verts per part, and how to determine the number of 8K texture maps per Reconstruction Zone such that even a beefy PC can sustain 60 fps? I know this needs to be tested, but many thanks for thoughtful responses guiding my testing and perhaps suggesting a formula to express the variables in play.