Avatar very noisy for a moment when it goes from occluded to visible #1038

Open
opened 2026-10-25 03:01:22 +00:00 by sky_line · 0 comments
Owner

When an avatar quickly passes from occluded (behind something) to visible, it looks very noisy for a moment before it settles.

Seen on the Vulkan renderer with RT GI and shadows. The owner doesn't think it's related to the GI kernel splits (#1031 / #1036); it's still under test.

Likely area: disocclusion. Newly revealed pixels have no valid temporal history, so the GI reservoirs / denoisers and the RT shadow denoiser restart from very few samples. Check what happens on history rejection for disoccluded pixels:

  • spatial fallback / wider filter until history builds up;
  • reservoir seeding from neighbours;
  • the moving-object (rigged) history-rejection path specifically;
  • whether TAA / FSR / DLSS also rejects there and adds to it.
When an avatar quickly passes from occluded (behind something) to visible, it looks very noisy for a moment before it settles. Seen on the Vulkan renderer with RT GI and shadows. The owner doesn't think it's related to the GI kernel splits (#1031 / #1036); it's still under test. Likely area: disocclusion. Newly revealed pixels have no valid temporal history, so the GI reservoirs / denoisers and the RT shadow denoiser restart from very few samples. Check what happens on history rejection for disoccluded pixels: - spatial fallback / wider filter until history builds up; - reservoir seeding from neighbours; - the moving-object (rigged) history-rejection path specifically; - whether TAA / FSR / DLSS also rejects there and adds to it.
sky_line added this to the Slipstream project 2026-10-25 03:01:22 +00:00
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
sky_line/Slipstream#1038
No description provided.