Shipping Crate

A showcase piece, not an example. Procedural crate (skids under the posts, corner posts with tenoned slats and seeded width jitter, bottom sills, nailed single-shell L-straps, filleted iron bail handles through mounting plates seated on one end board each) then the shipped pipeline: unique-cell UVs, Cycles high-to-low normal bake, LOD chain, convex collider, Unity glTF export.

Skids sit under the posts so the outer corner is a mortise, not a cave. Handle bails use quarter-circle corners so a 90-degree loft cannot sit the bar on the plate.

It asserts budget conformance of the generated result. It does not witness an API contract. "It rendered without error" is not a check.

Composes skills mesh-editing-and-bmesh, bake-high-to-low, depsgraph-and-evaluated-data, engine-export-presets, and snippets bake_normal_high_to_low.py, setup_bake_target_image.py, lod_chain.py / decimate_to_budget.py, convex_hull_collider.py, export_preset_unity.py (helpers copied, not imported as a package).

Budgets#

Declared as named constants; every gate recomputes from the mesh, materials, UVs, evaluated LOD, collider, or export file.

Base triangles rose from 624 to 2508 in the quality pass: the old beveled cube with glued-on slats became a post-and-slat crate with tenons, sills, L-straps, and pipe handles. The second pass took them to 3388: single-shell chamfered straps, 24 nails, chamfered mounting plates.

DECIMATE COLLAPSE triangle counts are not guaranteed identical across series; on this mesh they happen to match (LOD2 744 on all three). The gate is a ratio band, not an exact count. Bake pixels are stochastic; the gate is has_data plus operator FINISHED, not byte-identity. Slat-width jitter uses fixed seed 17. Export byte counts differ by 16 B on 5.2.1 (glTF serializer), not a gated axis.

Hygiene#

Recomputed from the generated mesh, not asserted about the script.

Joint fit and seat#

Second pass (straps, nails, handle placement, surface)#

The inspection sheet found the iron and the handles built the way crate-stack used to build them:

• Straps were two unbevelled boxes per corner. That left a seam line down the outer corner, razor edges, and an inner face sitting flush on the post face. Each strap is now one L-section extrusion that bites IRON_BITE into the post. It is chamfered at IRON_BEVEL with material=METAL_IDX, since bevel otherwise hands chamfer faces slot 0. Code copied from crate-stack, not imported.
• No fasteners. Three nail stations per plate, 24 nails, placed from each plate's own outer face and normal. nail_audit asserts the seat band (exit 18, --float-nails).
• The handle mounting plates straddled the gap between two end boards. The handle sat at a fixed 52% of post height, which landed on a gap. It now snaps to the centre of the end board nearest that height, from the same _span_layout the boards come from. plate_board_audit asserts each plate sits inside one board with clearance at both edges (exit 20, --straddle-handle, which puts the handle on a gap and measures −26.9 mm).
• Every plank was the same flat tone. PlankTone / GrainDir face attributes drive a grain-along-the-board wood shader. The iron gets rust wear.
• The temp glTF export is removed after it is measured.

The right-angle-edge budget from crate-stack is not applied here: the handle bail's pipe ends are square discs buried in the mounting plates, so "every box is chamfered" does not hold for this piece.

Third pass (stencil, hero angle)#

The crate read as a plain plank box: nothing on it said what it was for. Its front long side (-Y) now carries stencilled cargo marks, PORT ROYAL over NO 17. They are surface only: a 5x7 bitmap font painted into a small image covering the side in object space, cubic-filtered and thresholded into rounded stencil strokes, bitten by a fine overspray noise, and masked to the -Y faces. Each line is centred on one long slat, replayed from the same seeded _span_layout that places the slats, and its 64 mm cap height fits inside the board, so the baseline runs along the board. The first cut let both lines straddle a gap, which in perspective read as lettering sloping across the slats. The image node is deselected so the normal bake never targets it. Geometry, triangles, material face counts and every budget are unchanged; the glTF grows by about 800 bytes.

The hero turned the crate -48 degrees, which put the long side edge-on. The camera sits at about -51 degrees azimuth, so the crate now turns +15 degrees: the stencilled side is 24 degrees off the view axis and the handle end 66 degrees, a three-quarter view led by the lettering. An intermediate -20 degrees left the side 59 degrees off-axis, where the stencil foreshortened into a steep italic. The key is brighter and the stage is 60 m (the 14 m grid could show its edge).

Falsifiers#

Each violates one named budget. All nine were run on 4.5.11, 5.1.2 and 5.2.1 and returned the same code on each.

Run#

blender --background --python shipping_crate.py --
blender --background --python shipping_crate.py -- --skip-decimate
blender --background --python shipping_crate.py -- --stray-vert
blender --background --python shipping_crate.py -- --lift-z
blender --background --python shipping_crate.py -- --short-skids
blender --background --python shipping_crate.py -- --float-handle
blender --background --python shipping_crate.py -- --omit-slats
blender --background --python shipping_crate.py -- --float-nails
blender --background --python shipping_crate.py -- --straddle-handle
blender --background --python shipping_crate.py -- --output crate.png

Smoke passes no flags.

Exit codes#

File-local. 9 is a valid check code. 10 is reserved for gallery_framing.check_framing on the --output path. 15–19 are the hygiene and joint-fit family.