Rush

That brick wall in our kitchen? Well, we wanted to start with the first feature to highlight the white wall.

• ✓ Fakes geometric detail using texture data
• ✓ Light interacts with "bumps" that don't exist
• ✓ Zero extra polygons required
• ✓ Perfect for brick, stone, metal surfaces

We encode surface orientations into RGB values. At each ray hit, we sample this map, transform through the TBN matrix, and suddenly the lighting system believes it's hitting a bumpy surface.

The soft glow of the gas flame — this isn't point sources. They're shapes that emit light. Without area sampling, our renderer would waste countless rays hoping to randomly stumble upon them.

• ✓ (Should) Dramatically reduce noise
• ✓ Creates soft, realistic shadows
• ✓ Proper color bleeding between sources
• ✓ Essential for any scene with large lights

Trace a shadow ray, check for occlusion, compute the energy transfer. The math handles the rest — specifically, converting from area to solid angle measure.

Real cameras don't see clean edges around bright lights. The lens scatters. Bloom is our option to optical imperfection, creating soft lights and breaking hard edges.

• ✓ Creates cinematic, film-like quality
• ✓ Enhances perception of brightness
• ✓ Perfect for that cozy morning feel

Step 1: Extract bright pixels (threshold). Step 2: Blur them with Gaussian (separable for speed). Step 3: Add the soft glow back to the original. Multiple passes create that smooth, dreamy falloff.

Path tracing has a noise. Early iterations looks grainy. Intel's Open Image Denoise uses neural networks trained on millions of images to see through the chaos. It's a good solution for impatient renderers.

• ✓ Low sample renders become usable
• ✓ Render times drop dramatically
• ✓ Edges and details preserved

Feed OIDN three things: noisy image, surface normals, and albedo. Normals tell it where edges should be. Albedo separates texture from noise. The neural network handles the rest.