You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Description:
When using Volumetric Fog with Point Lights, there is significant temporal instability ("boiling noise") near high-frequency depth edges, such as alpha-thresholded vegetation. The noise pattern appears to crawl across the screen even when the camera and light sources are completely static.
Observed Behavior:
Visible jittering in the light volume.
The artifact is most pronounced where the light volume overlaps with complex depth (foliage).
The noise pattern cycles/moves continuously.
Proposed Solution / Feature Request:
Static Jitter Seed: Would it be possible to lock the jitter seed when the camera and scene are static? If there is no movement, the raymarching offset should remain constant to prevent "crawling" artifacts.
Depth-Aware Upsampling: The jitter is probably exacerbated by the upsampling process. Implementing a bilateral upsample (potentially using a 3x3 tent filter approach similar to the Jimenez/CoD Bloom implementation) might help the fog respect the edges of alpha-tested geometry better.
Karis Weighting: If the jitter is caused by high-intensity specular "fireflies" within the volume, a Karis-style weighting during the accumulation phase might stabilize the result.
Temporal reprojection: Maybe you already plan to address this with reprojection.
Description:
When using Volumetric Fog with Point Lights, there is significant temporal instability ("boiling noise") near high-frequency depth edges, such as alpha-thresholded vegetation. The noise pattern appears to crawl across the screen even when the camera and light sources are completely static.
Observed Behavior:
Proposed Solution / Feature Request:
Environment:
Attachments:
unknown_2026.04.20-15.47_clip_1.mp4