Ray-Guiding
08.03.2026 12:40
👍 2
🔁 0
💬 0
📌 0
Ray-Guiding
2048 sample
No mysterious Promotion and Decay rules.
No mysterious Promotion and Decay rules.
No mysterious Promotion and Decay rules.
No mysterious Promotion and Decay rules.
github.com/microsoft/Di...
hope this helps @okremiok.bsky.social c:
sw raytraced sky visibility with mainlight direction debug.
Using ReSTIR to importance sample the brightest pixels on an image
www.shadertoy.com/view/XfycWw
dynamic deform
soft body
UE5
ue5 approx soft body (non Niagara)
Lightweight Dynamic Deform System WIP
((‹( ˘꒳˘ )›))
((‹( ˘꒳˘ )›))
((‹( ˘꒳˘ )›))
UE5 dynamic deformation material WIP
Working on documenting frustum grids, rebuilding it piece by piece with Claude based on an earlier version that unfortunately got mangled by overly aggressive self-review: gist.github.com/vassvik/1ed6...
Still quite a bit missing, but getting there
IBL + dynamic sky occlusion shadow
Sky Visibility Debug on Surface Normal
sky occlusion ssprobe apply to IBL WIP
(゜-゜)RTX5070TI
Dynamic Sky Occlusion ( no HW RT)
(ﾟ∀ﾟ)ﾀﾞｲﾅﾐｯｸ
ScreenSpace Probe WIP
Fibonacci Spiral Points
テスト
www.shadertoy.com/view/w3tcWX
Current progress on my Bitonic sort obsession, compared against GPUSorting's OneSweep implementation (github.com/b0nes164/GPU...) as a reference
32-bit keys only. Tried to pick the best tuning parameters for the OneSweep reference to steelman the comparison
On my Blackwell Pro 5000 Mobile laptop
GTAO BentNormal (radius 32, 8*8*2=128 sample)
PlaneApproximate BentNormal (radius 32, 16 sample)
In HLSL, calling
GoldNoise(DispatchThreadID.xy, 1.0)
produces NaNs for certain values of DispatchThreadID.xy, e.g. (994, 581).
It doesn’t happen when I pass constant values, so it seems to occur only for non-uniform inputs ?🤔?
shadertoy.com/view/ltB3zD
low sample ao
referenced->
Jimenez et al. / Practical Real-Time Strategies for Accurate Indirect Occlusion
ハピニュ
readme update
Adaptive Software Tessellation with Concurrent Binary Trees
github.com/nagakagachi/...
dynamic Is All You Need
My "No Graphics API" blog post is live! Please repost :)
www.sebastianaaltonen.com/blog/no-grap...
I spend 1.5 years doing this. Full rewrite last summer and another partial rewrite last month. As Hemingway said: "First draft of everything is always shit".
Our #SIGGRAPHAsia 2025 paper "Sample Space Partitioning and Spatiotemporal Resampling for Specular Manifold Sampling" improves #realtime #caustics rendering.
Project Page: graphics.cs.utah.edu/research/pro...
@daqilin.bsky.social @tizian.bsky.social
"Neon"
for(float i,d,z;i++<1e2;){vec3 p=z*normalize(r.xyy-FC.rgb*2.)-1.;d=max(p.y,0.);p.y-=d+d;p.z+=t;z+=d=.3*(.1*d+.01+length(vec2(cos(p.z*6.)/6.,1.-dot(cos(p),sin(p).yzx)))/++d/d);o+=(cos(p.z+vec4(0,1,3,0))+1.5)/d/z;}o=tanh(o/8e2);
Left. The dominant axis is x. We first translate the ray along its direction vector such that the transformed ray’s origin o coincides with the subspace spanned by the other, non-dominant axes (only y in this 2D-example). Scaling the direction vector by 1/dx the x-component of the transformed ray’s direction d becomes 1. Now the intersection ray parameter with the near and far plane along the x-axis directly corresponds to the bounding box coordinates bminX and bmaxX without any further computations required
Axis-Normalized Ray-Box Intersection
"By transforming the ray we eliminate the need for one third of the traditional bounding-slab tests and achieve a speed enhancement of approximately 13.8% or 10.9%"
graphics.tu-bs.de/publications...
