# Crown Key material and prop starter kit R01

Production proposal for **Split Complex PI-BLD-0002-V01** and **Terrace Tower PI-BLD-0003-V01**, with a reuse map covering all 31 Crown Key catalog assets. Baseline main: `d88c6658de0439be2256b546e37bac03718b5ed6`.

**Delivered:** 13 machine-readable PBR material definitions, an editable vector swatch sheet, a deterministic Blender material-builder script, 24 reusable prop specifications, proposed dressing bills for seven spaces, 11 finish application records, and a reuse map for all 31 IDs. **Not delivered:** baked texture maps, modeled props, a `.blend` material library, automatic mesh assignments, or Blender/Unreal acceptance. The script is supplied for that next execution step; its bpy path has not run here.

## Use this packet

| File | Purpose |
|---|---|
| `material_library.json` | Authoritative proposed PBR values, sRGB and linear RGB, opacity/transmission/IOR, bump settings, tile size and density |
| `swatch_sheet.svg` | Editable SVG palette; schematic patterns, not PBR texture images |
| `swatch_sheet_preview.png` | Direct Inkscape rendering of that SVG for convenient visual review; not a texture map |
| `build_blender_materials.py` | Builds only named material datablocks; optional fresh material-only `.blend` library output |
| `prop_bill.json` | Proposed bounds, pivots, collision strategies, LOD budgets and material slots for 24 unmodeled props; unplaced quantities by space |
| `finish_application.json` | Split/Terrace space-specific finish assignments and upstream neutral-preview slot boundary |
| `family_reuse.json` | All 31 source IDs mapped to reusable existing material/prop families; variant overrides remain needed |
| `PROVENANCE.json` | Sources, observed hashes, inferred-versus-proposed distinctions and coordination record |
| `VALIDATION.json` | Actual checks completed and explicit unrun checks |
| `MANIFEST.json` | SHA-256 inventory of this packet, excluding the manifest itself |

The JSON values are editable sources. Run `python render_swatches.py` after color/value changes. Run `python validate_kit.py` before handing the packet onward. Reissue the manifest after any edit.

## Finish intent

The reviewed catalog and existing concepts support pale mineral structure, warm stone, restrained bronze, cool glazing, warm timber, neutral seating and tropical planting. Those visual families are source-backed. **The exact RGB/PBR values, tile sizes, density targets, prop bounds and budgets are new proposed authoring values**, not sampled material measurements or accepted geometry.

- **Split:** honed stone floors in lobby, gallery and receiving lounge; ivory/plaster structure; bronze mullions and trim; timber confined mainly to reception/furniture; pale linen seating. Use current `SP-L19-GALLERY` and `SP-L19-RECEIVING` IDs at the lead's proposed 88 m datum, not the superseded L20 suggestion. Preserve gallery central strip `y=-2..2 m`, receiving arrival strip `x=20..29 m` and the lead's glass-edge circulation. Roof planting remains scenic/nonplayable.
- **Terrace:** honed stone lobby/lounge, timber studio floor/cabinetry, warm timber lounge/terrace soffit accents, exterior stone on the flush garden terrace. The studio bath subzone uses stone/ceramic pending its actual partition/clearance review. Use the lead's `L00-*` IDs; the closed core approach does not gain an open lift interior.
- **Other variants:** `family_reuse.json` preserves the existing R02 use/room program. It proposes kit compatibility, never a new scene assignment. Crown Landmark issued finishes remain controlling and are not automatically replaced.

The atlas has no named scene assignment for Split or Terrace. Do not add council/vote/diplomacy props to them merely because those scenes exist elsewhere in Crown Key.

## Color, glass and scale contract

`base_color.srgb_hex` and `srgb_rgb` are display-referred authoring values. `linear_rgb` is computed using the standard sRGB transfer function and is what the script sends to Blender numeric shader inputs. Future color/emission image files use sRGB; roughness, metallic, normal, height and AO maps use non-color data. Do not double-convert an sRGB image.

Glazing uses proposed IOR 1.5, roughness 0.055, transmission 1 and **alpha 1**. Physical transmission and cutout opacity are different controls. The proposed 12 mm glass thickness is a modeling input to review, not a construction specification. Refraction, reflections, sorting and the actual Unreal material mode remain untested. Foliage is an opaque leaf-geometry starter; no alpha card, species or wind rig is delivered.

Object coordinates are local metres and require applied object scale 1. The procedural mapping divides coordinates by `tile_size_m`; each future UV0 tile spans that physical size. A 2 m tile with a proposed 1024 px map gives 512 px/m; a 1 m tile gives 1024 px/m. Analytic metal/glass/light has no texture-map budget yet. Lightmap UVs, streaming, atlas packing and actual runtime budgets remain separate work.

Procedural bump values are distance in metres plus a unitless strength. They add microvariation only. The timber swatch's illustrated lines are **not** a delivered grain texture. The Blender noise shader is also not authored wood grain. Create or source the real texture/geometry pass when the selected room needs it. Warm light emission is a unitless preview multiplier, not lumens or validated photometry.

## Blender execution handoff

The authoring environment has Python 3.12.14 but no `blender` executable or `bpy`. The following Python-only check was run successfully:

```bash
python build_blender_materials.py --validate-only
```

In a Blender environment, use a fresh output path:

```bash
blender --background --python build_blender_materials.py -- --output crown-key-materials-r01.blend
```

The script creates only the 13 `CK_MAT_*` datablocks. It does not edit scene units, meshes, slots, routes or the user's original scene. With `--output`, it calls `bpy.data.libraries.write` with only the created material datablocks, rather than saving the current scene; it refuses to overwrite an existing output. It also refuses to replace a same-named material that lacks this kit's ownership marker. Runtime socket lookup supports known Principled transmission/emission naming alternatives and fails visibly for missing required inputs; actual version compatibility still requires execution.

Run this in a staging file, inspect a stone/metal/glass/foliage sample under neutral and warm lighting, then record actual Blender version and evidence. Procedural node graphs are not engine-ready texture maps. Bake or rebuild the chosen functions and test the exported asset in Quik's actual Unreal version before claiming matching appearance.

Upstream OBJ previews use neutral `wall`, `slab`, `solid`, `core`, `glass` slots. **Do not globally replace every wall/slab with one finish.** The geometry owner must assign explicit component/polygon slots or finish meshes. A single slab can need a stone top, mineral edge and timber soffit. This kit intentionally leaves those assignments unapplied.

## Prop production rules

Every prop is a specification with a proposed bounding box, not a modeled asset. All quantities are initial dressing estimates with null transforms. Place them only after checking the owning route, door swing, player/camera envelope and window clearance. No universal clearance value is imposed by this packet.

Use +Z up, +Y front, metres and scale 1. Floor-centered props sit on local Z=0. Wall-mounted/back-aligned props use a rear-bottom pivot; the door uses a hinge root and separate frame/leaf collision. Foliage is nonblocking and sits in a separately blocking planter. Simple convex collision and LOD triangle/screen-size values are proposals to test, not measured performance claims. Runtime geometry, UVs, normals, lightmaps, materials and LODs remain production tasks.

The common lounge, table, planter and lighting families can serve several buildings after fit review. Office workstations, large meeting tables, commercial cafe equipment and specific booth gear are still missing; the all-31 reuse map flags the need rather than relabeling unrelated props as finished replacements.

## Checks completed

`VALIDATION.json` records successful JSON/cross-reference checks, sRGB conversion checks, PBR range/alpha checks, tile-density arithmetic, prop bounds/LOD ordering, room finish references, current Split IDs, all-31 source-ID mapping, Python AST parsing and SVG XML parsing. A static source check confirms the material-library writer is present and whole-scene saving is absent; this is not a Blender execution test. Regenerating the SVG produced identical text. Inkscape rendered the SVG preview; the rendered sheet was visually inspected for legibility and clipping.

Blender material creation, rendering/baking/export, Unreal import/shading, furnished fit/collision/traversal and LOD/performance checks were **not run**. No engine acceptance, actual prop geometry or automatic spatial-fit approval is claimed.
