Hacker News new | past | comments | ask | show | jobs | submit login

The article makes the point that it became inefficient to use displacement maps due to their size ballooning beyond '36 times' what they would normally use. So this means that the ground displacement map was just not detailed enough and pushing it to higher resolutions didn't scale. Instead they used Houdini to generate sand grains, I presume Houdini used some sort of displacement map per grain? The final result is a mesh nevertheless.

They also mention that, "this result was driven by a combination of culling techniques including camera frustum, facing angles and distance, creating a variance of dense to coarse patches of sand for optimum efficiency."

So while they had grains each of ~5000 polygons available to them and positions that they could occupy, they cut down on that polygon count using a combination of techniques (many common in games too) to make it all work. So the polygons of any grains outside the camera frustum were ignored. The faces of the grains point away from, I presume, any light source and therefor not contributing to the path tracer were ignored and also some soft of LOD to reduce polygon count as distance.

As an amateur blender user I'm reminded of microdisplacement as a way for us mere mortals to achieve something in this space: https://www.blenderguru.com/tutorials/introduction-microdisp...

Doesn’t aggressive culling present issues with ray traced lighting? If you remove an umbrella that’s behind the frustum then a shadow in the scene is missing. Even harder is if you cull a second layer of sand grains then you don’t get the light that reflects off of them (assuming partial transparency of grains).

Yes, so I presume they only remove the faces that are not exposed to any light bounce.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact