aae1ebdcfd
4 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
jing
|
a99f733452 |
[lane A] MODELBEAST dresses the molecules: 5 matcaps, and the canal picks space-filling
The right axis wasn't "generate or don't" — it was WHICH HALF. Geometry stays procedural because a molecule's shape is already known exactly. **Material is exactly a generation problem**, and it's the whole difference between a textbook diagram and a game object. So MODELBEAST made the materials. 5 greyscale matcaps, flux_local, 69 seconds, $0.00, 68 KB total: mol_glass (O,N) — the big win: oxygen is a red glass marble with internal glow mol_matte (C) — graphite soot; carbon reads as the scaffold it is mol_chrome (Co) — B12's cobalt is now a chrome bearing in a cage mol_molten (P) — ATP's phosphate tail is incandescent AND pulses (uTime, free) mol_pearl (H) — satin white; hydrogens stop shouting The trick is Lane D's own law applied to spheres: author the LUMINANCE, tint in-shader. derive_maps already greyscales every matcap, so one glass ball serves oxygen AND nitrogen at their own CPK hues — 5 balls dress the whole periodic table we care about. They pack into one strip atlas indexed per-vertex by aMat, so **a molecule is still exactly 1 draw call** however many materials it's made of (7 molecules = 7 draws, unchanged). ?matcap=0 falls back to the built-in fake-lit path: the assets-optional law holds. SPACE-FILLING WINS FOR PICKUPS, and the canal decided it, not me. New `repr: 'space-filling'` draws every atom at its real van der Waals radius with no sticks. A/B'd at real pickup size against the real L2 wall, same camera: ball-and-stick goes spindly and dissolves; the solid blob holds its silhouette — and it's CHEAPER (21k vs 26k tris, no bond cylinders). Recommendation to B: space-filling in flight, ball-stick for hero/UI/collect close-ups where the chemistry is worth reading. Evidence: round2_molecule_materials.png (3-way). Two defects found by rendering it, both mine, both fixed: - Tint x luminance darkens TWICE: CPK carbon is 0.4 and a matte ball averages 0.5, so soot x soot vanished — the first pass sank glucose, caffeine and the cobalt into the background. Floored the matcap at 0.22, the same floor the fallback's key light always had. - The molten matcap was my own bad prompt: I asked for an "incandescent white hot CORE" and got exactly that — a black ball with a hot spot, so phosphorus lost its orange. Re-rolled for the whole sphere to glow (attempt 2 of the <=2 PIPELINE allows). Lesson for the kit, next to the organ/tissue one: a matcap is a LUMINANCE LOOKUP, so a dark-dominant ball makes a dark-dominant object. Also corrected the doc's own framing — v1 said "I'm authorised to burn GPU and I'm not going to", which was the wrong axis, not just the wrong tone. qa GREEN; 16/16 texture provenance verified (synced the FIXED batch json up first this time, so the box couldn't clobber it on the way back). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
jing
|
3a6dbc30a3 |
[lane A] Complete the biome texture set (oral + colon) + a provenance gate
Generated on MODELBEAST flux_local, on-device, $0.00. Lane D's pipeline used
unchanged; **D owns these — ratify, retune or bin them.**
Why A did D's job: the six-biome tint contact sheet could only judge 3 of 6
biomes for real (oral/large_intestine/appendix had no pack and fell back to
procedural), and you cannot ratify a tint against a stand-in. All six are now
textured — the set ART_BIBLE §Biome palettes promises.
- wall_oral_a/b, wall_colon_a/b + normals. Seams 6.35/9.30/14.06/6.37 -> 1.01/
1.08/0.74/1.36 (D's "indistinguishable" band; the shipped pack's weakest is
2.39). Means normalised to ~0.50 by D's derive_maps, untouched.
- appendix wears the colon pack at its own gold tint (TEXTURE_SHARE in
world/index.js) — anatomically it IS colon tissue, and it's ART_BIBLE's
"one texture, two biomes" law for zero bytes. Distinct from the temporary
slug map; keep it when that dies.
- oral_a took three subjects. v1 named an ORGAN ("tongue") and FLUX drew the
organ's silhouette — D's "one urchin of villi" in a field's clothing. v2 named
the tissue but papillae are 3D projections, so perspective gave it a vanishing
point that tiles into starbursts. v3 asks for squamous cell pavement: flat by
construction. Rule for the kit: prompt the SURFACE, not the FEATURE.
Evidence: docs/shots/laneA/round2_oral_prompt_evolution.png
PROVENANCE DEFECT FOUND AND FIXED (-> Lane D):
batch_textures.json recorded, for 4 of 6 walls, the prompt from the framing
experiment D tried and rejected — not the prompt that made the shipped pack.
record() only runs on generation, so reverting a prompt while keeping its image
(exactly what D correctly did: "four winners kept untouched rather than churned")
drifts the record silently. Proved by regenerating esophagus_a at seed 1101 under
both prompts: the code's prompt reproduces the shipped wall, the recorded one
gives the bland wood-grain D described rejecting. Evidence:
docs/shots/laneA/round2_provenance_probe.png. Records re-pointed at the code's
prompt (images untouched — they were never wrong).
New: gen_textures.py --verify-provenance + qa.sh gate 5b, so it cannot rot again.
qa GREEN incl. the new gate; 11/11 provenance verified; all six biomes eyeballed
in engine (docs/shots/laneA/round2_tint_*.png).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
||
|
jing
|
efea7d98f3 |
[lane D] Round 1: first texture pack + audio proof, exact-tiling pipeline
Textures (6 walls + normals + wet matcap, $0, ~64s on the m3ultra MPS): - FLUX ignores "seamless tiling" (seam_error 3.7-17.9); a 4-way cosine partition-of-unity blend of the four half-rolls fixes it exactly -> 0.87-1.32 - levels-normalised the pack to one exposure (means were 0.18-0.32); these are detail maps multiplied into a biome tint, so per-prompt exposure read as a broken tint rather than as dark tissue - prompt-kit experiment rejected + recorded: PROCITY's "uniform coverage" clause in the STEM cured centred subjects but flattened the pack; framing words now live only in the two subject prompts that needed them Audio (1 bed + 4 sfx, 4.2s render, 0.61/10MB, ogg+m4a dual-ship): - bed-esophagus 24.000s, loop seam 0.32 (wrap step 3x smaller than inner step) - cascaded lowpass poles: one-pole at 900Hz left 16kHz only ~25dB down = hiss - rewrote the spectrogram sheet (per-clip peak, log freq) — the first one was a saturated rectangle, and evidence you can't read is not evidence assets.js: miss ledger + misses() — Lane A found that a drifted slug falls back procedurally forever with no error. Each distinct miss now announces itself. Also: shot_sink.py (canvas -> docs/shots/laneD, correctly named) and a dev texture viewer that imports the stub world read-only for the eyeball law. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |
||
|
jing
|
dade036f69 |
[lane D] Round 1: pipeline scripts + assets.js loader
Ports PROCITY's pipeline shapes to GUTS (dry-run first, idempotent, resumable): - gen_textures.py: flux_local wall/matcap pack, ART_BIBLE prompts, provenance JSON - derive_maps.py: exact-tiling (4-way cosine partition-of-unity) + Sobel normals + WebP - build_manifest.py / validate_manifest.py: catalogue-what-exists + qa hook - gen_audio.py: procedural numpy synth, seamless loops, ogg+m4a dual-ship - js/core/assets.js: manifest loader, null-on-miss, ?localassets=0 empty boot Generation runs next on the m3ultra box. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |