68 lines
3.6 KiB
Markdown
68 lines
3.6 KiB
Markdown
# Aligned Half-Life office walls and ceiling
|
|
|
|
## Research and asset choice
|
|
|
|
Perplexity was unavailable. Exa found the community VERC guide and original-game
|
|
material catalog; actual candidates were then inspected from the locally installed
|
|
Half-Life `halflife.wad` (not textures from the Black Mesa remake).
|
|
|
|
- https://twhl.info/index.php/wiki/page/VERC%3A_Style_Guide_1
|
|
- https://greg-kennedy.com/hl_materials/
|
|
- https://twhl.info/index.php/vault/view/6388
|
|
|
|
The VERC guide identifies green `C1A0_LABW*` wall panels and `C1A0_C1` ceiling in
|
|
the early laboratory aesthetic. For the furnished OFFICE room we chose the original
|
|
FIFTIES family instead: `FIFTIES_WALL14U` (128x160, plaster above wood paneling) and
|
|
`FIFTIES_CEIL5` (48x48, four square acoustic panels). The guide is a community
|
|
reference, not proof that this exact pair appeared in a specific original office.
|
|
Candidates were compared visually before selection. Original pixels are preserved.
|
|
|
|
`tools/extract_office_surfaces.py` exports just these two textures to
|
|
`assets/textures/office/`, plus source WAD and texture-lump SHA256 manifest. It
|
|
never changes Steam files or re-extracts zombie assets.
|
|
|
|
```sh
|
|
python3 tools/extract_office_surfaces.py /path/to/Half-Life/valve/halflife.wad
|
|
```
|
|
|
|
## UV layout
|
|
|
|
The old generic world-space triplanar mapping did not distinguish wall motifs from
|
|
ceiling motifs. `room.gd` now retains BoxMesh vertices/normals/indices but rebuilds
|
|
wall/ceiling meshes as ArrayMesh with explicit room-space UVs. Existing collision
|
|
boxes and placement remain untouched.
|
|
|
|
- Walls: U advances clockwise along the room perimeter at one texture per 4m.
|
|
V runs floor-to-ceiling once, inverted so the wood is BELOW the plaster. Equal
|
|
density128px/4m =160px/5m, no aspect distortion or repeated vertical trim bands.
|
|
All window segments share the same coordinate origin; segment size does not
|
|
reset the mapping. The 72m perimeter is18 full repeats, so all corners close.
|
|
V uses a half-texel inset at the floor/ceiling to avoid exact0/1 wrap-edge sampling.
|
|
- Ceiling: XZ planar UVs with a shared origin at(-9,-9). One48x48 image covers1.2m
|
|
and contains two panels per axis, so each panel is0.6m. The18m ceiling spans15x15
|
|
texture repeats (30x30 panels). No per-face rotations or arbitrary local scaling.
|
|
- Linear mipmapped anisotropic filtering and U tiling remain enabled.
|
|
- Floor keeps its existing texture/projection. Furniture, lighting and gameplay
|
|
are unchanged. The old wall/ceiling PNGs are retained but no longer used here.
|
|
|
|
## Cross-check with the GoldSrc mapping project
|
|
|
|
The shared memory already recorded, for the same FIFTIES family in the `goldsrc-mapping`
|
|
project on x280, two conventions: a room's wall height should coincide with the wall
|
|
texture height so there is exactly one repetition without cuts, and textures carrying a
|
|
baseboard band must be kept upright with the texture base at the floor. The choice above
|
|
(one vertical repetition, inverted V so the wood sits at the floor, 32 px/m on both axes)
|
|
applies the same rule independently. That project is not a source of instructions here;
|
|
it corroborates the numbers already derived from the texture dimensions.
|
|
|
|
## Verification
|
|
|
|
`tools/test_office_uv.gd`: seventeen explicit-UV surfaces, no triplanar on office
|
|
materials, shared UVs across window segments,4m wall period, wall corner phase
|
|
continuity/full-perimeter closure, and ceiling15x15 tiling. PASS headless and real
|
|
GL Compatibility. Images `/tmp/zombie-office-uv-{0,1,2}.png` inspect opposite walls
|
|
and ceiling; first two use a TEMPORARY inspection light, last uses actual gameplay
|
|
lighting. The inspection light exists only in the test, not in the game.
|
|
|
|
Existing resource-in-use warnings at test shutdown remain.
|