ASSET_BUILD_LIST.md

assets/ASSET_BUILD_LIST.md

The asset requirement — derived from canon, not from registry population

Rebuilt 2026-08-13. The previous version of this file reported rows currently in the CSVs as

though they were the build target. They are not. They are how far the authoring cascade has got.

Every registry below is a partially-populated layer, and several are 1-3% populated. The

requirement comes from the structural canon — 79 nodes, 76 regions, 22 threads, 10 trades × 12

gates, 72 weapons, 34 equip slots, 7 item grades, 67 factions, 33 legendary species, 185 abilities

— and the arithmetic is shown for every line so it can be argued with instead of trusted.

The one distinction that makes this buildable: an ITEM is not a MESH, and a SPECIES is not a

MESH. Seven rarity grades re-treat one base form. A juvenile, an adult and an alpha share one

skeleton. Regional armour variants are material and trim on a shared silhouette. Tens of thousands

of items come off low thousands of meshes, and that ratio is the whole production plan.

---

1. The requirement, per category

Locked numbers that are already correct and must never inflate: **72 weapons · 185 abilities ·

22/33 familiars · 33 legendary species · 79 nodes · 22 threads · 12 side questlines · 120 trade

quest gates.** Everything else above is a floor, not a ceiling.

---

2. Where each number comes from

Bosses — ~500

284 rows today. Gap ≈ 215. Note that 149 of the 284 are unrouted and 42 creature-side rows are

blocked on a body-plan ruling, so the *usable* figure today is lower than 284.

Creatures — ~2,000

A region needs an ecology, not a bestiary entry. Per region: 1 apex predator, 3 mid predators,

5 herbivores, 4 birds, 3 aquatic, 4 insect/small, 3 domestic/livestock, 2 mounts, 1 scavenger

= 26 species.

• 76 regions × 26 = 1,976
• 11 realms × ~10 native fauna = 110
• folklore beings per region beyond the ecology (the golden-standard enrichment: real malevolent

beings from the culture's own tradition, plus its fairy-tale lore) ≈ 5 × 76 = 380, of which

roughly half double as boss-class

Species ≈ 2,100–2,400. Base meshes ≈ 2,000. Skeletons ≈ 40 once a body-plan taxonomy exists

(quadruped-digitigrade, quadruped-unguligrade, biped-avian, serpentine, arachnid, cephalopod,

insectoid-hexapod, and so on). 187 rows today. Gap ≈ 1,900.

The 42 rows tagged NEEDS_CANON_BODY_PLAN plus 7 PROVISIONAL are the leading edge of this, not an

edge case — at 2,000 species the body-plan taxonomy is the single highest-leverage canon artifact

in the whole 3D lane, because it decides how many skeletons and rig sets exist.

Named characters — ~1,500

167 rows today. Gap ≈ 1,320. Of the 1,500, only the fairy-realm ~80 carry pointed ears.

Background NPCs — ~7,000

Standing canon: every entity has a persona, a life, a job and relationships, and killing one ripples

through the community. That is a *data* requirement at full scale and a *mesh* requirement only at

archetype scale.

• 76 regions × ~3 settlements × ~30 populated inhabitants = 6,840

Meshes required: ~120. Roughly 8 archetype bodies per region cluster, driven by MetaHuman with

per-NPC variation (face blend, skin tone, groom, outfit swap, height/weight). This is precisely the

category MetaHuman exists to serve, and it is the difference between 7,000 sculpts and 120.

Antagonists — ~240

• House of Velheim operations across 79 nodes, ~3 operatives per operation = 237
• plus the named family tree and the institutional brokers/complicit states through which §17.1

routes the story antagonist

64 rows today. Gap ≈ 175.

Equipment — ~30,000 items off ~2,400 meshes

The multiplication, stated so the mesh cost is visible:

• 34 equip slots (T0_Equip_Slot_Registry), of which ~14 are visible-on-body armour slots and

~20 are jewellery/trinket/seat slots

• 7 item grades — common, uncommon, rare, epic, legendary, mythic, signature

(T0_Rarity_Grade, item_grade axis). Grades are a visual treatment on a base form, never a

new mesh

• 76 regional material cultures — a Sámi coat and a Khmer one are different silhouettes; a

bracelet's material changes, not its form

Mesh arithmetic: 14 visible armour slots × ~30 regional armour vocabularies = 420 armour meshes;

20 small slots × ~40 distinct forms = 800 small meshes; consumables, tools, trade goods,

reagents, quest items ≈ 1,200. Total ~2,400 meshes.

Item arithmetic: 2,400 forms × 7 grades × material tiers ≈ 30,000+ item entries, before the

gear-imprint rune lane, which is sigma-bounded stat rolls on an existing item and costs zero

mesh.

23 rows today. This registry is effectively unbuilt, and it is the largest single authoring gap

in the corpus.

Weapons — 72 forms, ~6,000 states

72 is LOCKED (36 Brute + 36 Trickery/Sorcery) and correct. But a weapon is not one asset:

• Lane B transformative upgrades land on the twelve quest gates per capability, and a

transformative upgrade that does not change the weapon's appearance is a wasted gate

• the Proficiency Level 8 Mastery ability and Level 99 Legendary transformation are named canonical

Lane B instances

• LEGENDARY and MYTHIC are rarity grades opened by the final mastery questline

72 forms × 12 mastery states = 864 appearance states; × grades and imprint treatments ≈

6,000 item entries. Mesh work: 72 hero forms + ~200 variant/upgrade meshes. Cheapest volume

in the project per asset, no identity problem — the right place to prove the pipeline.

Vril sites — ~230

Per the combat addendum §14, a vril site is the **save point, the vril recharge, and the loadout

sanctum**. That makes density a level-design requirement rather than a lore count: a region with

three vril sites is a region with three save points.

76 regions × 3 = 228, plus the 6 Vimana coordinates and the realm-side sites.

42 rows today. Gap ≈ 190.

Interiors and flora

• Interiors: 76 regions × ~12 enterable structures (dwellings, workshops, shrines, stores,

faction halls) = ~900. 44 rows today, and T0_Dwelling_Registry holds 2.

• Flora: 76 regions × ~12 biome plants + herbalism reagents across 10 trades = ~1,200.

25 rows today.

---

3. What this means for the split of work

Meshes Josh actually has to build: ~7,000–8,000, not 905. Broken down by how they get made:

The number that matters is not the total — it is that **~5,700 of these have no identity bar at

all**. Equipment, flora, environment and interiors are pure throughput work where a good result is

a good result. Only the ~2,100 character-and-boss meshes carry a likeness constraint, and only 7 of

those are the protagonist.

Rows I have to author before any of it can be built: ~13,000+, against ~1,100 mesh-bearing rows

that exist. That is my side of the split and it is the real bottleneck — a mesh cannot be built for

a creature that has no design sheet, and 1,900 creatures have none.

---

4. Priority order

1. The creature body-plan taxonomy. Not 42 blocked rows — the ruling that decides how many

skeletons exist across 2,000 species. Highest leverage artifact in the lane.

2. CHAR_0001, seven rungs. Everything in the slice stages behind the protagonist existing.

3. Weapons, 72 forms. Cheapest per asset, no identity problem, proves the pipeline on volume.

4. The slice creatures and bosses (Prologue → Ch 13), against a populated roster rather than

the 152/52 that current rows happen to contain.

5. One MetaHuman end-to-end, before 1,500 rows queue behind an unmeasured cloud round-trip.

The browser Creator sunsets 2026-11-05; the in-engine route is the survivor.

---

5. What I build for you, per asset

For every row above, from the canon rather than from chat:

• The design sheet — the nine-block bestiary grammar. 109 exist; the format is proven and

critic-gated.

• The conditioning prompt set — front / left-profile / back, dimensional rather than

adjectival. Every place I wrote "smaller" or "low-rise" instead of "two inches below the navel",

the model ignored it.

• The region page it belongs to, and the canon cites that justify its design.
• Care-tier and hard-line flags where §17.1 applies.

6. Standing rules for the plates I hand you

Learned expensively over 2026-08-12/13:

• Every plate is matted before it conditions anything. Nothing in the old chain matted the

input, and the model invented backgrounds — brick walls, in the measured case — because it had to

guess what was subject.

• Cut at the neck, never at a fixed height, and use the same ring on both halves.
• Views must agree on where the subject ends. Eight views that crop the body at 1.48 to 2.35

face-heights below the chin produce a torn chest, because the reconstruction tears where the

views disagree.

• Garment spec is dimensional. Waistband two inches below the navel, five inch inseam. A base

doll's garment is a liability: whatever it covers becomes body geometry made of cloth, and every

equipment layer inherits it.

Generated by harness/site/structure_site.py — the URL path is the repo path. review root