Unreal 5.1: Problem with Google photorealistic 3d tiles when ignoring KHR_materials_unlit

Cesium for Unreal
1

I am working on a project where having realistic scenes is very important and we were very excited about the updates for Google photorealistic tilesets and especially the recent update which meant we’d be able to have lighting interacting with the terrain.

The tiles and lighting look great, however there seems to be a bit of an issue with the surface normals that becomes evident when there are large amount of tiles which are meant to be aligned on a relatively flat surface, such as fields or bodies of water. The tiles essentially appear to not line up nicely and I’ve not found any way (ie, ticking generate smooth normals) of convincingly eliminating this effect.

Ideally it’d be nice to have the option to use a water mask in the case of the example I show here, but this example was more one where it’s super evident - as above it’s really any “flat” area that can show this problem. Is there any work around, or likelihood of a fix for this to be coming soon?

Cheers

image
image1439×681 64.9 KB

2

Have you turned on the “IgnoreKhrMaterialsUnlit” property here? They’re really not intended to be rendered with the “unlit” extension ignored, and so you basically get what you get when you turn on that property.

Unfortunately the Google Photorealistic 3D Tiles don’t work great with lighting currently, because they don’t include normals. We can generate them (and that’s what we do when you check that box), but we can’t distinguish the real geometry from the skirts that are meant to hide the seams between tiles. Because the skirts are vertical, if their normals are generated in the usual way they end up with much different lighting from the nearby geometry.

If I have all that right, you should provide the feedback to Google that you would like the tiles to include normals so that they work better with lighting.

3

Yeah I’ve turned it on, that’s what I mean by “Ignoring KHR_Materials_Unlit” :slight_smile: And ok, that makes sense and we’re on the same page but I’m winding up the wrong people to try and get a resolution to it. Hopefully more people find this and make noise as it feels like it’s going to be a common complaint when more people start trying to use the Google tiles.

Cheers.

4

Yep, sorry, I missed that in the title!

We’ve mentioned the lack of normals to Google as well, and hope they will add them in the future (maybe as an option?). They understandably can’t solve every problem at once, particularly with a massive dataset like this that surely has a lot of inertia!