Chapter 3: Texture / Material Budgets & Atlases

Created by Sarah Choi (prompt writer using ChatGPT)

Texture / Material Budgets & Atlases

Optimization, Modularity & Reuse for Prop Concept Artists (Prefabs, Kits, LOD Thinking)

Texture and material decisions are downstream costs you can influence at the concept stage. The way you cluster color breaks, define trim zones, and plan decals determines how many materials, how many textures, and how much memory the final prop will consume. This article equips both concept and production prop artists to design with budgets, atlases, and trim sheets in mind—so kits scale, LODs degrade gracefully, and your scene stays within performance targets across platforms.

1) Budgets 101: what actually costs

Real‑time cost is a mix of GPU memory, bandwidth (texture fetches), and draw calls/material switches. As a rule:

  • Fewer materials per prop → fewer draw calls.
  • Shared atlases/trim sheets → better batching/instancing.
  • Coherent texel density → predictable LOD and mip behavior.
  • Right compression → more art in less memory, with acceptable artifacts.

Your concepts should therefore minimize unique paint islands that force extra materials and encourage reuse of trims and decals rather than bespoke textures.

2) Texel density planning (concept‑level)

Define texel targets before painting:

  • Baseline: e.g., 512 px/m for general props, 1024 px/m for hero faces, 256 px/m for background set dressing.
  • Continuity: keep kit parts within ±10% to avoid LOD pop and shader aliasing.
  • Priority zones: mark hero read zones (logos, dials) that warrant higher density or unique tiles; mark trim zones for shared edges and ribs.

In the concept pack, annotate scale bars on orthos and a texel density chip for quick checks.

3) Trim sheets vs unique textures vs decal atlases

Trim Sheets (shared edges + materials)

  • Best for: repeating edges, ribs, bevel lips, screws, gaskets.
  • Concept guidance: draw trim lanes directly on callouts (e.g., 8 cm bevel, 12 cm rib, rubber gasket) so modeling/UVs align with the architected trims.
  • LOD: trims survive distance; detail normals remain stable; easy to reuse across a kit.

Unique Textures (hero plates)

  • Best for: large emblem faces, narrative stickers, bespoke paint wear.
  • Concept guidance: limit the count and size; isolate to one material set if possible.
  • LOD: provide a reduced macro variant for far LODs (less text, thicker marks).

Decal Atlases (stickers/labels/glyphs)

  • Best for: serials, barcodes, hazard icons, inspection stamps, tape tears.
  • Concept guidance: deliver vector masters + two scales (micro/macro) of the same iconography and placement rules (no crossing seams; keyline included).
  • LOD: at distance swap to bold glyphs with thicker strokes.

4) Material count & instancing (the batching lever)

  • Target: ≤2–3 materials per prop (body + accents + decals). Whole kits should share one master material with parameter sets (color, roughness, grime, emissive).
  • Instancing: consistent material assignment across the kit enables hardware instancing. Concepts should keep palette choices within that master so production doesn’t split materials.
  • Channel packing: design for ORM (Occlusion/Roughness/Metallic) or similar packing to cut texture reads.

Annotate on the concept sheet: “MasterMaterial_MM_Generic_PBR; Params: BodyColor, HazardColor, Wear, Grime, EmissiveNits.”

5) Compression & platform notes (design to avoid artifacts)

  • Normals: author as BC5/ATI2 (two‑channel) where possible to avoid blocky artifacts; avoid fine diagonal noise patterns that macro‑block.
  • Basecolor/ORM: BC1/BC7 on desktop/console; ASTC 6×6 or 8×8 on mobile/VR.
  • Alpha: only when necessary (decals, cutouts). Alpha textures compress worse; prefer mesh splits or packed masks over full alphas when design allows.
  • High‑frequency patterns (mesh/cloth) should be detail normals + tilables, not baked unique, to survive compression.

Concept implication: avoid ultra‑fine printed micro‑type and hairline stripes that will die under compression; show a macro alternative for LOD.

6) Building an atlas that stays stable

Atlas design principles

  • Group by function: labels/hazard, grime overlays, numerals, and icon glyphs in separate quadrants.
  • Padding & mips: 8–16 px padding between elements at final atlas resolution to prevent mip bleeding.
  • Stroke safety: minimum 3–5 px stroke at atlas res; include white/black keylines where contrast is low.
  • Orientation: align glyphs to consistent axes; avoid rotated UV islands when possible to improve compression.

LOD & variant strategy

  • Provide A (micro) and B (macro) variants of the same symbols on the atlas. At distance, shader swaps the UV region.
  • For number/letter serials, supply modular sheets (0–9, A–Z) and encourage blueprint‑driven composition, not unique textures.

7) Detail normals, tilables, and masks (the texture stack)

  • Detail Normals: small‑scale noise layered in shader for up‑close richness; switch off at L1/L2 to prevent shimmer.
  • Tiling Materials: paint/steel/rubber authored as tileables applied via trims; keep tilling frequency consistent across kit (document the target tiling/m).
  • Mask Overlays: wear/grime/paint chips as grayscale masks multiplied in shader; expose wear amount as a parameter for variety without new textures.

In concepts, provide a Material Matrix with tiling targets, roughness ranges, and where masks should accumulate (top edges, under fasteners).

8) LOD thinking for textures (not just meshes)

Design the texture simplification ladder alongside geometry:

  • L0: full decal specificity, micro text, high‑frequency normal details.
  • L1: drop micro text; replace fine patterns with broader fields; halve detail normal strength.
  • L2: keep only silhouette‑critical color blocks, one glyph, soft normals; switch emissive strips to pips/impostor cards.

Include a Distance Read Board with 100%/33%/12.5% scale crops to simulate mip behavior.

9) Prefabs & kits: how budgeting shapes the design

  • Shared atlas per kit: one Hazard_Atlas and one Decal_Atlas span the entire kit so level artists don’t create unique materials per prefab.
  • Trim sheet per family: e.g., Industrial_Paint_Metal_Trim; add rubber/ceramic lanes if the kit needs it.
  • Hero exceptions: a small Unique_Hero texture set reserved for marketing or interactable props. The concept sheet should gate these exceptions explicitly.

Provide a Budget Table: Expected materials, texture sets, and approximate memory (at target compression) for L0/L1/L2.

10) Color, value, and readability under compression

  • Value separation first: choose body values in N4–N6 so hazard bands and decals have headroom.
  • Keylines: add white/black pinstripes around low‑contrast edges to resist bloom and mip bleed.
  • Color‑blind safety: pair hue with shape + numerals on the atlas; never rely on hue alone.

Concept implication: include grayscale thumbnails of the paint scheme and a color‑blind preview to prove readability regardless of compression shifts.

11) VR/AR specifics (mobile budgets)

  • Compression: prefer ASTC 8×8/6×6; avoid uncompressed alpha where possible.
  • Angular size: ensure decals/glyphs maintain ≥0.4° visual angle; scale via world‑space in shader.
  • Material count: be ruthless—1–2 materials per prop; share trims across the scene.
  • Temporal comfort: emissive pulses ≤3 Hz sustained; no aggressive flicker patterns that exacerbate reprojection artifacts.

Design angle: simplify atlas stroke weights and avoid near‑Nyquist patterns that shimmer in HMDs.

12) Marketing (key art and thumbs) texture pass

  • Produce a Marketing Variant of the atlas with thicker strokes (×1.25–1.5) and simplified labels for store thumbnails and trailers.
  • For hero renders, consider temporary uncompressed or higher‑res atlas to avoid JPEG artifacts; keep a game‑legal variant for in‑engine.

Document this as a separate export profile so teams don’t ship the marketing textures.

13) Case mini‑studies

A. Storage Crate Family

  • Design: trims for ribs/edges, shared Decal_Atlas (serials/hazard), unique 512² hero plate for emblem.
  • Result: one master material + parameters; memory per crate under 3–4 MB at L0; clean L1/L2 swaps.

B. Pipe & Conduit Kit

  • Design: tileable metal/rubber trims, decal arrows/flow codes on atlas; detail normals off at distance.
  • Result: zero unique sets; batchable across long runs; no shimmer.

C. Bio‑Sample Canister (Hero)

  • Design: one unique 2k set for cap + body (marketing), plus shared hazard atlas. L1 macro glyphs; L2 emissive pip impostor.
  • Result: hero looks rich up close; gameplay and trailers use simplified atlas without new textures.

14) Naming & handoff

  • Materials: MM_Prop_Generic_PBR, MM_Prop_Trim_Industrial, MM_Prop_DecalAtlas.
  • Textures: TX_Trim_Industrial_Base_2k_BC7, TX_Atlas_HazardGlyphs_1k_BC1, TX_Trim_Industrial_Norm_2k_BC5, TX_Trim_Industrial_ORM_2k_BC1.
  • Parameters: BodyColor, AccentColor, WearAmt, GrimeAmt, EmissiveNits, DetailNormalScale.

Include this block exactly on the concept page to prevent proliferation of ad‑hoc materials.

15) Closing

Great props feel rich not because they hoard textures, but because they stack smart materials, trims, decals, and masks within a tight budget. When you design with atlases and texel density in mind—and you plan a texture LOD ladder that preserves the message—your prefabs and kits become endlessly reusable, performant, and consistent across FPP, TPP, isometric, VR/AR, and marketing contexts. Treat texture and material planning as part of the concept, and you’ll save memory, time, and headaches all the way to shipping.