Lightnews — Scholar-powered news

Reposted by Brian Karis

themaister.bsky.social @themaister.bsky.social · 4d

New blog post: Outlining an alternative path to learn graphics programming.

themaister.net/blog/2025/10...

A case for learning GPU programming with a compute-first mindset – Maister's Graphics Adventures

themaister.net

10 30

Reposted by Brian Karis

Arseny Kapoulkine @zeux.io · 9d

New blog post! In "Billions of triangles in minutes" we'll walk through hierarchical cluster level of detail generation of, well, billions of triangles in minutes. Reposts welcome!

zeux.io/2025/09/30/b...

1 54 160

Reposted by Brian Karis

Dylan Browne 🔜 GCAP @dylanbrowne.bsky.social · Jul 24

The new Nanite Foliage Voxel representation is an absolute game changer for distant environments, density preservation is pretty much perfect.
5 min test scene i whipped up here.
~1 million 20 million poly trees scattered with PCG
#UE5 #UnrealEngine #GameDev #EnvironmentArt #RealTime

6 14 72

Reposted by Brian Karis

c0de517e.bsky.social @c0de517e.bsky.social · Jul 14

My talk is out! c0de517e.com/023_hpg.htm

Hallucinations on the future of real-time rendering.

Angelo Pesce's homepage & blog on computers, graphics and other things.

c0de517e.com

6 15

Reposted by Brian Karis

Christoph Peters @momentsingraphics.bsky.social · Jun 21

All HPG 2025 papers are now available on the EG digital library.
CGF papers: diglib.eg.org/handle/10.23...
Conference papers: diglib.eg.org/handle/10.23...
It has been a huge honor to act as HPG papers chair and guest editor and I look forward to an exciting program in Copenhagen.

44-Issue 8

High-Performance Graphics 2025

diglib.eg.org

1 24 52

Reposted by Brian Karis

Justin Solomon @justinmsolomon.bsky.social · Jun 18

I signed and encourage colleagues to do the same.

Venue selection is fraught. Inevitably attendees will encounter visa issues, distance, expense, local politics, and other challenges. This venue, however, is simply unsafe for LGBTQ+ researchers, and the leadership's defensiveness is unacceptable.

Omer Shapira @omershapira.com · Jun 18

Computer Graphics as a field avoids gatekeeping. Practitioners come from many disciplines, origins and beliefs.
SIGGRAPH has historically carried this practice.

I am thus outraged at the decision to hold SIGGRAPH Asia 26’ at a venue unsafe to many.

Adding my signature. I encourage you to do too:

Open Letter to the SIGGRAPH Leadership

RE: Call for SIGGRAPH Asia to relocate from Malaysia and commit to a venue selection process that safeguards LGBTQ+ and other at-risk communities. To the SIGGRAPH Leadership: SIGGRAPH Executive Commit...

docs.google.com

1 8 23

Reposted by Brian Karis

Matthew Wilde @skilful.bsky.social · Jun 13

This is a great thread @hillsguy.bsky.social, I'll add a bit more context here for anyone interested.

Yes there's no fancy sim - it just uses a water interaction buffer as a target for the particle system rendering, so we could use the power of particle system we already have.

Hillsguy @hillsguy.bsky.social · Jun 11

Decided to check out how water is rendered in Counter-Strike 2.

Given how interactive water in CS2 is, I expected to see some cellular fluid simulation, but turns out they just use a low-resolution off-screen buffer with flat 2D decals used for various aspects of water. Reminds me of HL2 water :)

4 52 240

Reposted by Brian Karis

Stephen Hill @selfshadow.com · Jun 10

The SIGGRAPH Physically Based Shading course is back this year: blog.selfshadow.com/publications...

SIGGRAPH 2025 Course: Physically Based Shading in Theory and Practice

blog.selfshadow.com

1 30 81

Brian Karis @briankaris.bsky.social · Jun 5

Yes but not ready to share anything yet

1 4

Brian Karis @briankaris.bsky.social · Jun 4

Thanks Sean!

Brian Karis @briankaris.bsky.social · Jun 4

Includes this great video showing how the Nanite Foliage seamlessly LODs. No popping. No harsh transitions. Good appearance preservation (although this light direction is overly generous).

1 27

Brian Karis @briankaris.bsky.social · Jun 4

More rendering tech details on how the Witcher 4 UE5 demo works youtu.be/0X6amtHcrUE?...

Livestream 2, Day 2 | Unreal Fest 2025

YouTube video by Unreal Engine

youtu.be

1 5 39

Brian Karis @briankaris.bsky.social · Jun 4

1 minor detail, it can switch to voxels within the twigs themselves as they get further as well. A Nanite cluster can now be either 128 triangles or 128 voxel bricks. That visualization only showed 5 steps but there are as many as it takes to halve the count until there is 1 root cluster.

2 18

Brian Karis @briankaris.bsky.social · Jun 4

Great video from DF!

Really good explanation from @dachsjaeger.bsky.social about how foliage has been done in the past, the problems with them, and why this new approach was needed.

Digital Foundry @digitalfoundry.bsky.social · Jun 4

The Witcher 4 tech demo running on Unreal Engine 5 - 60fps on the standard PlayStation 5. Alex and I share some initial thoughts: youtu.be/5PIyPpB83Vw

3 5 37

Brian Karis @briankaris.bsky.social · Jun 4

I’ve wondered the same. We haven’t yet explored any alternative uses for these new tools yet. It probably can’t beat the best hand modeled hair cards but it seems like it could be great as an auto generated LOD for spline based grooms.

3

Brian Karis @briankaris.bsky.social · Jun 4

Yes. This has been one of the most challenging parts of the problem. If opacity per voxel is too expensive, voxels can either be 100% covered or not, which do I fill? And determine that relatively fast since the Nanite build for a mesh needs to take <1min.

1 1

Brian Karis @briankaris.bsky.social · Jun 4

A cluster in Nanite typically is up to 128 triangles. Voxel clusters are 128 4x4x4 bricks. Brick positions are fixed point so no precision problems.

Really though gaps aren’t a key concern for voxels since they are used primarily for subpixel features which result in partial coverage.

4

Brian Karis @briankaris.bsky.social · Jun 4

Correct. 3 separate features: assemblies, skeletal meshes, and voxels. None of which is strictly foliage. Each can be used without the other. Can make assemblies and keep as meshes, like the repeated elements on this shield from the original UE5 demo. Or voxelize a single Nanite mesh.

6

Brian Karis @briankaris.bsky.social · Jun 3

Sure. We didn't get time to do the wind animation as an actual physics sim but want to in the future. AC Shadows wind looks amazing and we aren't there yet.

Breaking off branches could be a problem since Nanite assembly HLOD needs to merge things. They aren't individual branches in the distance.

2

Brian Karis @briankaris.bsky.social · Jun 3

Agreed

Brian Karis @briankaris.bsky.social · Jun 3

5.7 for first version and then probably more significant improvements for 5.8.

1

Reposted by Brian Karis

Yining Karl Li @yiningkarlli.bsky.social · Jun 3

The new Witcher 4 UE5.6 tech demo that was shown today is the most amazing realtime tech demo I've seen yet. Nanite Foliage is particularly impressive- I've always liked the idea of voxel-based LOD and this is the best implementation I've ever seen!

www.youtube.com/watch?v=aorR...

The Witcher 4 Unreal Engine 5 Tech Demo 4K | State of Unreal | Unreal Fest Orlando

YouTube video by Unreal Engine

www.youtube.com

3 21

Brian Karis @briankaris.bsky.social · Jun 3

We have a USD (which supports nested instancing) importer. CDPR had their own custom speedtree exporter/importer using our Nanite Assembly API for these assets. We have tooling in the works that will help with authoring in the future. Can't talk about it yet.

2

Brian Karis @briankaris.bsky.social · Jun 3

Not fixed. No voxel alignment between any cluster, including parent/child. Like the triangle clusters it targets a halving of data. Half the triangles is easy, just collapse until hit. Voxels need a search since voxelization needs to be done from scratch for each size attempt.

1 4

Brian Karis @briankaris.bsky.social · Jun 3

Using a simplified version of this citeseerx.ist.psu.edu/document?rep.... Basically compute the extents of deformation using only the extents of the skinning data contained within a node. This gets too expensive once the bone count within a node gets high at which point we fallback to the object.

citeseerx.ist.psu.edu

4