Commit 4dcb601
fix(showcase): remodel pine-tree as a dense, plated-bark mature Scots pine
The pine was the weakest nature piece. It sat small in the frame, its
crown was a sparse stack of bottlebrush shoots that read as holly, and the
branch skeleton showed through. It is rebuilt as a mature Scots pine:
- a taller bare bole (to ~3 m) with twelve deep plate ridges fading out
between 1.9 and 3.0 m, and bark shifting from grey-brown plates to
fox-red over 1.9-3.4 m, so the upper bole and every limb are red
- eight whorls of 3-5 heavy crooked limbs (was ten of 4-6), long and near
level low down and shorter and steeper above, so the crown is broad and
rounded (9.3 x 7.9 m) with a flattened top
- 113 needle brushes. Each is one closed shell: a small dark cube-sphere
core (3x3 a side) whose 108 split facets are each drawn out into one
long, thin tapered strand. Strands aim outward plus a hashed scatter, so
they splay through a dome, not a plane. Every limb, one to three side
twigs per limb and the leader end inside their own brush. A Tip point
attribute shades each strand from base to point, and a NeedleUV map
lays a line down each strand. The soft blue-green tone varies per
brush, lighter outside and high. Strands are stylised sprays (~0.4-0.8
m long, a few cm wide); botanical 1-2 mm needles would need ~10x the
triangle budget.
- five dead stubs and two longer, thicker dead limbs on the bole
- a needle-litter soil disc (humus with rust and grey needle streaks):
the trunk foot and five roots are bedded in it, with five fallen cones,
two mossy cleaved stones and two dead sticks lying on it
- six materials (bark, needles, cone, deadwood, soil, stone)
- relit: key 3000 -> 5200 W, fill 170 -> 520 W, wedge 1900 -> 2250 W and
slightly more saturated, world 0.02 -> 0.052, so luma is in the
calibration band
Two foliage drafts were rejected in review. Pyramid-faceted pads read as
crystal shingles. Wide-facet pom-poms read as flat star rosettes (holly or
maple at thumbnail). The small core with scattered thin strands replaced
both.
New budgets and falsifiers:
- Clump seat (exit 22): every brush's own carrier, tagged via a Carry face
attribute, has its deepest vertex >= 0.040 m inside the brush
(ray-parity signed depth; measured 0.0592-0.1366, 113 brushes).
--short-twigs stops every carrier at 0.64 of its pad's radius, same
point count: shallowest -0.1955, exit 22.
- Ground bedding (exit 23): cones and sticks 0.004-0.040 m under the soil
(measured ~0.010), stones 0.015-0.120 (0.045), roots >= 0.030 (0.091),
trunk foot >= 0.050 (0.1175). --float-litter lifts the cover 0.08 m:
exit 23.
- Checks 22 and 23 run before the component check (21), so a falsifier
that also detaches parts still names its own budget.
- The asset-quality remap moves from 22 to 24.
Budget changes:
- Base tris 32500-36000 -> 43500-46000 (measured 45852; LOD 0.5000/0.2200)
- Outer AABB -> (9.284, 7.932, 7.659) m; DBH 0.266 -> 0.377 (0.3768)
- Whorls 10 -> 8, limbs 3-5 per whorl, gap band 0.40-0.68 -> 0.32-0.60
(measured 0.4037-0.4479); 33 limbs + 7 dead members seated 0.4259-0.4656
- Lean 0.489 deg, balance 0.0466 m (bands unchanged)
- Collider hull from a 14-sided trunk: 56 tris (20 sides gave 58-62,
10 gave 80) under the unchanged ceiling of 60
- --lean-crown now bends only the bole below the lowest whorl and carries
the crown over rigidly. The old whole-axis bend grew the X extent
17.7 mm and exited 8; the bend was moved, no band was widened.
Falsifiers (flag -> measured -> exit), all on the final code:
--skip-decimate LOD1 1.0000 9
--stray-vert loose_v 1 15
--lift-z zmin 0.05000 16
--float-branches seat ratio 1.2447 17
--lean-crown lean 1.205 deg, envelope unchanged 19
--bunch-whorls 7 tiers, gap 0.7121 20
--drop-cones 10 components 21
--short-twigs brush bite -0.1955 22
--float-litter cones -0.0705, stones -0.0352 23
tests/check_falsifier_targets.py --run --only pine-tree passes.
Evidence (live runs, local):
- .scratch/5.2/blender-5.2.1-windows-x64/blender.exe (Blender 5.2.1 LTS).
Two default runs exit 0 with identical measured lines; hygiene all 0
including coplanar cross-shell pairs; one connected assembly of 266
shells.
- Framing: fill x 0.509 (was ~0.33), y 0.794; margins l .247 r .244
b .083 t .122.
- Contact sheet: luma 0.2399 (before 0.193; band 0.237-0.243), stage
0.2508, sat 0.1546 (band 0.1525-0.2367), warmth +0.153 (the old hero
was +0.12; both are warmer than the calibration set).
- Asset sheet: the tree now reads as a pine of needle brushes over red
limbs on a soil disc. It is not the least-designed object in the lineup;
the hydrant is simpler.
- 4.5 LTS and 5.1 were not run locally.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>1 parent 460477a commit 4dcb601
11 files changed
Lines changed: 2335 additions & 993 deletions
File tree
- docs/gallery
- asset-sheets
- assets
- contact-sheets
- pine-tree
- showcase
- pine-tree
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
146 | 146 | | |
147 | 147 | | |
148 | 148 | | |
149 | | - | |
| 149 | + | |
150 | 150 | | |
151 | 151 | | |
152 | 152 | | |
| |||
241 | 241 | | |
242 | 242 | | |
243 | 243 | | |
244 | | - | |
| 244 | + | |
245 | 245 | | |
246 | 246 | | |
247 | 247 | | |
| |||
Large diffs are not rendered by default.
Loading
Loading
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1433 | 1433 | | |
1434 | 1434 | | |
1435 | 1435 | | |
1436 | | - | |
| 1436 | + | |
1437 | 1437 | | |
1438 | 1438 | | |
1439 | 1439 | | |
1440 | | - | |
1441 | | - | |
| 1440 | + | |
| 1441 | + | |
1442 | 1442 | | |
1443 | 1443 | | |
1444 | 1444 | | |
| |||
0 commit comments