According to the docs, the specular value should be left at 0.5 for almost all materials. Everything else looks correct, only the light source reflections look strange.
Can you take a video of the camera moving to make the stripe transition more visible? Also, what does the PT output?
When the camera moves horizontally or zooms in and out, the stripes also change. Slight adjustments to the camera angle will change the direction of the stripes. With a more significant change in the camera angle, the stripes will disappear. This issue does not occur in offline rendering.
Is there any plan for native support for Restir GI in Lumen?
The prototype hasn’t been updated for a year, and I wonder if the idea has been completely abandoned or if there are plans to revisit and complete it.
Also, for reflections, will there be an option to recalculate accurate reflections with proper shading and high-quality GI in the future?
1 Like
restir again? what even is this? reading papers…
restir is basicly temporally and spatially accumulated lighting. isn’t that what taa and the motion and sample history does? and everybody hates it, apparently.
sure… lumen uses irradiance probes instead of pathtracing, which makes it a lil coarser, but also compatible with lower end hardware. it runs in dx11. console level. hmm
until the next gen comes out with raytracing features restir can wait. a lil bit. that’s my opinion. i’m just a game dev. yo
1 Like
Nvidia is already working on integrating ReSTIR GI into their branch as an additional option for developers using Lumen hardware, replacing the probes in Lumen. Epic had started a prototype to further enhance Lumen, but it hasn’t been updated in a long time. I was wondering if Epic plans to revisit this project and whether they will eventually offer a more advanced method for recalculating reflections to ensure they are accurate, without relying on the surface cache.
2 Likes
I’ll respond but put it in blocks so as to not throw the thread more off-topic:
Summary
I think Epic is unlikely to implement a renderer like this for real-time anytime soon. I’m familiar with the paper, and the performance numbers don’ currently look doable for real-time, regardless of the graphical quality.
Furthermore, iridescent effects like scales, soap bubbles, metal, etc. are all things that substrate makes it quite possible to convincingly fake. In particular, the SpecularLUT node lets you create a very convincing illusion of thin-film interference when deployed properly, and I don’t think that kind of physical accuracy is as big a priority for UE as performance is. Not to say I wouldn’t love more PBR capabilities in-engine, just that I don’t think it’s likely anytime soon.
1 Like
@Krzysztof.N There’s a lumen bug in 5.4 that I’m attempting to understand (LMK if it’s already been fixed):
In the safe house content pack, I keep receiving circular rendering artifacts in the lumen scene on these metal pipes. The normals don’t appear too unusual to me, and I can’t figure out any obvious reason for the errors. Lumen scene detail is 3, lumen scene lighting is 8, hit lighting enabled, normal proxy geometry mode (although nanite geometry also demonstrates the artifact).
This sheet also demonstrates rendering artifacts, but it includes this fragmented appearance in the reflections view for no discernible reason. Enabling/disabling WPO on the cloth makes no difference whatsoever, and both sides of the material exhibit these artifacts. The PT renders the scene as expected.
Is the pipe hallow? That and the sheet might be having artifacts due to being thin.
Does it happen without Nanite? Try r.Nanite 0. There’s a long standing bug with Lumen Surface Cache Capture + Nanite + Two-Sided materials.
Yes, we do plan to deliver a higher quality Lumen targeting high-end PC GPUs. Hard to say at this point what will be the specific solution or when it will happen, as we need to spend some time figuring it out.
11 Likes
I tested it both ways, Nanite and non-nanite. Some objects had their rendering issues clear up, but others persisted:
Is it possible to achieve realistic mirror reflections using lumen
With Lumen reflections and Lumen GI, in most cases, no.
With Lumen reflections and Lightmapped GI, yes.
1 Like
Actually appears I’m wrong, in 5.4.3 Lumen reflections with lightmapped GI seems to be broken. It’s only sampling the volumetric lightmap for lighting… I guess the answer is no for both cases if you’re on 5.4.3 
2 Likes
I haven’t worked with lightmapping in a while, how can you tell the static geo is only sampling the volumetric lightmap? The angular resolution of lighting?
The most significant indicator is that the VLM stores lighting in a very sparse grid and just interpolates to the nearest probe without any occlusion, so it leaks light through walls. Low spatial resolution also means it can’t capture any meaningful lighting detail such as the indirect shadows from the sphere/pillar on the floor.
You can’t see the rest of the scene but there is light leaking in the reflections on all of the walls, and the ceiling.
Here’s a much more obvious example:
xcuse me, but… what is this whole shabang doing there in the first place? it’s a wall. that doesn’t make sense. that’s not a leak but a wrong mirror.
a pretty scuffed render. i dunno.
eh I think it’s just reflecting something offscreen and I drew the arrow to the wrong place
The geometry in the reflection is definitely correct, it’s just the lighting that is messed up.
video
1 Like
