TAA: unjitter the current-frame taps (sparkle slice 2, −22% static flicker)#114
Conversation
The resolve sampled the current frame at its JITTERED position (sample_catmull_rom(in.uv), no offset), so alpha x jitter-phase difference shimmered into every output frame by construction. The current tap and its 3x3 clamp neighborhood now resample at the true position via TaaParams.yz (= -0.5*jitter_ndc.x, +0.5*jitter_ndc.y in composed UV - the GL-convention divide flips the sign, and the variance rig arbitrated it empirically). History reprojection stays in jittered space on purpose: EN-022's velocity-ref VP already bakes the current jitter. Measured (f12-live rig, 5 live captures, high-pass temporal stddev): bank 3.99 -> 3.11 grass 3.46 -> 2.70 (taa-off floor 1.94/1.03) Combined with slice 1 (motion-aware disocclusion edge): -22% static flicker. Gates: sharpness A/B clean at rs 1.0; rs 0.75 title capture sharp, no ghosting (TSR super-resolution intact - the sub-pixel information now enters through the Catmull-Rom reconstruction instead of leaking out as output noise). Remaining gap to the floor is NOT TAA: an in-run indoor capture caught the artifact the user actually reported - white render-pixel-scale square speckles blanketing interior surfaces (GI-probe-texel shaped, SSGI prime suspect). Archived in the shooter as tools/.testout/indoor-blizzard.png; separate hunt.
|
Warning Review limit reached
Next review available in: 9 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
1 participant
The resolve sampled the current frame at its jittered position — α × jitter-phase-difference shimmered into every output frame by construction. Current tap + clamp neighborhood now resample at the true position (
TaaParams.yz; sign arbitrated empirically by the variance rig). History reprojection deliberately stays in jittered space (EN-022 velocity-ref bakes the jitter).Measured: bank 3.99→3.11, grass 3.46→2.70 (with slice 1: −22% static flicker). Gates: sharpness A/B clean at rs 1.0; rs 0.75 title sharp, no ghosting — the jitter's sub-pixel information now enters via Catmull-Rom reconstruction instead of leaking out as noise.
Also: an in-run indoor capture caught the user's actual report — white GI-probe-texel-shaped speckles blanketing interiors (archived
tools/.testout/indoor-blizzard.png). That's a separate SSGI-suspect hunt, now the top open item.