Granite: red lath preset tuning

Branch jiadeng/tbricks-semantic. Renders: Cycles, 32 samples, head-on camera, 1024 px, rendered on this Mac (mymba). Newest first. Earlier rounds and generator results are on the previous page: pumini.tailf92817.ts.net/granite. Click any image for full size.

Red lath preset tuning, round 3: fine matrix grain, crisper flecks, fewer white crystals (2026-10-13)

Pink fixed at the preset's. Variants tune_z1-tune_z4 in lath_tune_materials.py, each adding one change to the previous one; the preset itself is unchanged. Rendered locally (Metal), 1024 px.

share of pixels by sRGB lightness band (tone_classes.py; median color where useful)
              L<0.2    0.2-0.35   0.35-0.55   0.55-0.75   >0.75
photo crop     4.4%     12.3%      26.0%       46.9%       10.4% (.87,.81,.76)
preset        12.2%      7.9%      23.7%       48.9%        7.2% (.79,.78,.76)
y1             5.4%      7.7%      31.0%       48.6%        7.3% (.80,.78,.76)
z1             4.8%      8.2%      31.4%       48.3%        7.3% (.80,.78,.76)
z2             4.9%      8.5%      31.3%       48.1%        7.3% (.80,.78,.76)
z3             6.0%     10.5%      37.0%       43.7%        2.8% (.81,.77,.74)
z4             5.3%      9.3%      34.4%       47.7%        3.4% (.80,.76,.74)

• z1: fine crisp matrix grain is back without the black/white speckle (darkest band 4.8%, photo 4.4%). But the matrix is now a uniform fine salt-and-pepper everywhere; the photo's matrix gathers into larger charcoal patches and separate granular light-gray patches.
• z2: fleck softness 0.03 is hardly distinguishable from 0.06 at this size.
• z3: fewer, smaller, warmer white crystals bring the mid-dark band closest to the photo (10.5% vs 12.3%), but the brightest band falls to 2.8% (photo 10.4%): the white crystals were most of the render's brightest pixels, and the mid band overshoots (37% vs 26%).
• z4: the granular light grains hardly show at this share.

Reference photo