We called these "Lumen Virtual Lights". With MegaLights they probably won't see much use in future projects though. :)

That's correct. It works surprisingly well and didn't really degrade performance that much. :)

Absolutely!

I agree. This is mostly why I didn't stay in academia to come back to the industry. :)

I'm glad that my little visualization helped to illustrate how distance fields work. :)

I do want to stress though that the morphs in keeper do not use meshing or in other words when objects combine / blend / separate no new mesh is computed for rendering, mostly because it would be prohibitively slow for the resolution we need. I'll talk about how we render the morph surfaces soon.

You can see this nicely in this closeup. Notice how the number and orientation of the edges changes on the left side as both meshes are merged.

The topology of my madness is a infinitely regressing manifold! Haha.

For the implicit surface? This is done using the distance field gradient.

Hehe. Clay... the original implicit surface representation. :)

Of course implicit surfaces representations have other problem. I'll talk about these issues in the future.

Hehe. Indeed. :)

Thanks. Yeah I’m planning to talk through a bunch of stuff going forward. :)

Here is the sphere filling test I did with the intention to only use it for shadows: bsky.app/profile/p1xe...

You are right. In fact our shadows can be leaky at times especially since we are using virtual shadow maps. All in all it works well enough for us. At the beginning I experimented with an alternative representation for shadows but it never worked well enough.

Yes we aren’t meshing the iso surface. I was inspired by @iquilezles.bsky.social awesome articles on SDF modeling and wanted to see how far we can push it w/o mesh generation. Turns out pretty far. :)

It’s fun! :)

And there is some pretty wild fan fiction as well. Haha.

It’s actually a game. I’m just sharing some information about some of the tech. :) bsky.app/profile/doub...