Chapter 1: Texture / Material Budgets & Atlases

Created by Sarah Choi (prompt writer using ChatGPT)

Texture / Material Budgets & Atlases for Character Concept Artists

Why optimization and accessibility start at the sketch

Texture and material choices decide not only frame rate and memory but also who can comfortably read the design. A jacket with three clearcoats and a 4K unique texture can be breathtaking in a key art render yet muddy or strobing in gameplay, especially for players with low‑vision or color‑vision differences. Concept artists set the slope of these outcomes. When you plan budgets, atlas schemes, and contrast logic in the concept stage, you help production ship characters that are both fast and inclusive—clear at distance, legible on SDR monitors, and friendly to a wide range of players and platforms.

Budgets you directly influence

The parts you draw determine how many materials a model needs, how many texture pages it touches, and how stable those materials will be under mip‑mapping and compression. Think in three linked envelopes: material count per slot, texture page count and sizes, and texel density per surface class. Heads, hair, outfits, palettes, and decals each carry a micro‑budget. If every slot follows its envelope and you provide graceful fallbacks, teams can mix and match variants without memory spikes or shader stalls.

Atlas thinking as a design language

An atlas is not just a technical map—it is a style scaffold. By deciding which elements live together on a sheet, you define which details can be reused across variants and how trims carry identity when resolution falls. Group soft goods (undersuit knits, lining fabrics, elastic bindings) on a Soft Kit Atlas with a harmonized texel density and shared roughness logic. Place armor plates, buckles, and hard trims on a Hard Kit Atlas with stronger micro‑normal and controlled specular ranges. Keep hair on a Hair Atlas that supports masked transparency without contaminating other slots. Reserve a Decal Atlas for emblems, numbers, and typographic elements with stable edges, and manage palettes with either a small palette LUT plus mask maps or a tint mask atlas that can be shared across the set.

Material caps as creative rails

Material limits protect both performance and clarity. Treat them as rails you compose within rather than fences to fight. A production‑friendly guideline is one opaque PBR material per soft garment, one or two materials for a hero hard piece (base plus a restrained emissive), one masked material for hair, and a single decal material shared across the character. If a hood needs transparent fabric, first explore an opaque lace motif baked into albedo and normal before granting true transparency. When you must exceed a cap, trade somewhere else so the total per character remains stable. Make these trades visible on the board so tech art can predict draw‑call behavior.

Texel density for readability and fairness

Texel density should follow read priority, not ornament cravings. Faces need the highest stable density to preserve eye and mouth landmarks; hands deserve enough density that finger silhouettes and glove seams survive the gameplay camera; torsos and capes can taper. Establish a density ladder in your art notes and keep it consistent across variants. This not only simplifies baking and LODs, it supports fair competitive reads: enemy heads remain crisp enough to read intent, friendly class markers do not blur into the background, and decals maintain minimum letter heights that survive mips.

Mip‑safe detail and anti‑moiré thinking

High‑frequency textiles and micro‑patterns are beautiful in closeups and hazardous at distance. Draw patterns with scalable motifs: blocky repeat units that compress gracefully into tonal fields. Where you need tight weaves or pinstripes, provide a far‑LOD variant on the same atlas—larger repeats or simplified fills that the material can swap to beyond a screen‑size threshold. Avoid fine checkerboards, tight herringbones, and sub‑pixel stripe widths that will shimmer under motion. If a tartan is core to identity, pin its dominant bands to value‑separated rails so the silhouette remains readable when hue information collapses.

Compression, color space, and inclusive contrast

Textures will be compressed, and not all players sit on wide‑gamut displays. Paint with compression resilience in mind: keep large smooth ramps (skin, painted enamel) free of banding by limiting unnecessary gradients; move subtle variation into low‑amplitude noise that survives BC1/BC3 while avoiding blotches. Separate metalness/roughness/AO into channel‑packed masks to reduce pages, and choose contrast that remains legible in SDR with a standard gamma. If your palette relies on red/green opposition to signal class, provide a shape or value‑based carrier so color‑vision diverse players receive the cue. For emissives, set the design intent to accent, not flood; blooms that obliterate neighboring form can be painful for light‑sensitive players and waste HDR headroom.

Palettes as data, not textures

Colorways don’t need a new albedo per outfit. Favor a palette LUT and mask maps that remap channels at runtime. In concept, show one neutral render and several channel remap plates with the same lighting to prove that class reads and faction identity hold across dyes. Explicitly protect skin and hair by keeping them off tint channels unless the fiction demands otherwise; this avoids palette swaps that push complexions into unreadable or culturally insensitive spaces. Keep value rails for each channel to prevent mid‑tone mush where edges and decals disappear.

Decals that scale and localize

Decals are text and symbols first, textures second. Design them for resolution elasticity: a crisp vector‑friendly shape at 1k that still reads as a solid icon at 128 px. Favor bold glyphs and modest inner detail; outline strokes improve survival under motion blur and TAA. Plan legal zones on relatively flat surfaces to reduce stretching, and avoid seam crossings and high‑curvature arcs that distort letters. Where player text is permitted, set minimum sizes, max character counts, and fallback fonts with full localization coverage. Place all decals on a shared sheet and reuse them across sets; this not only saves memory, it keeps faction language consistent.

Hair without overdraw explosions

Hair eats fill‑rate and can dither into mush under compression. Design styles as clumps and ribbons rather than clouds of flyaways. Keep masked transparency on a single material and design shapes that convert to ribbons and then a cap texture for distant LODs. Reserve true transparency for thin bangs only where cinematic value demands it. Colored tips and emissive streaks should be thick enough to survive mip reduction; if a gradient is core to the read, anchor it in albedo and normal rather than relying solely on a translucent overlay.

Heads that survive mips and still emote

Faces require stable shading and careful makeup/facepaint strategies. Prefer opaque eyes at gameplay distances with baked highlights and AO to avoid refractive complexity. Eyeliner, freckles, and tattoos should have two renditions: a close‑range version with subtle noise and a far‑range version with simplified edges and higher contrast. Facial hair should migrate from cards to textured geometry blocks at LOD2 to prevent alpha chatter. Build all of these paths into your concept plates so model and shader authors can match intent.

Layering without page creep

Layered outfits invite material inflation. To stay within budget, plan style carriers that are cheap: stitched trim bands, panel breakups, and embossed fields that bake to normal maps. When you need a glossy accent, consider a mask‑driven clearcoat baked into the roughness rather than a separate clearcoat layer. If the design calls for translucent visor or veil moments, capture their opaque fallback for gameplay use while reserving the transmission look for portraits or cinematics. Mark these boundaries explicitly on the board so downstream teams can hook them to camera context.

Streaming and UI swap stability

Players will preview cosmetics in menus and swap pieces quickly. Design with atlas reuse to minimize streaming hitches: keep undersuits and common trims on stable pages that rarely change, and move unique hero moments into small overlays. Propose capsule swaps—tied‑back hair, short cape wedge—so set changes don’t force entire wardrobe reloads. Any time your concept introduces a new atlas, ask what it replaces or whether that page can be shared across two or more characters.

Accessibility checks you can run without a build

You can catch many readability issues at concept time. Downscale your plate to the target gameplay pixel sizes and check whether class markers, facial landmarks, and decal text survive. Convert to grayscale to ensure shape and value carry the read independent of hue. Try a quick color‑vision simulator pass mentally or via a reference to confirm that red/green or blue/yellow dependencies are backed by secondary cues. Dim the image to simulate low‑contrast displays and verify that silhouettes and critical trims still separate. If a read fails any of these checks, adjust palette value spacing, thicken accent bands, or enlarge legal‑zone decals.

Documentation that keeps budgets honest

Attach a small budget box to each board summarizing expected texture sheets, sizes, materials, and known overdraw risks, plus a note on palette strategy and decal legal‑zones. Provide an atlas plan diagram with arrows from each garment piece to its sheet. Include a revision log when you move content between atlases so tech art and rendering can track why streaming behavior changed. Most importantly, keep your naming consistent with the slot taxonomy so automated tools can harvest your intent.

Example envelopes you can tailor to your project

Numbers vary by engine and platform, but you can sketch envelopes now and adjust later. A head often targets one skin material with a 2K shared atlas entry plus a 1K eye sheet; hair uses a single masked material on a 1K page; a chest hero piece may justify a 2K sheet if it serves multiple sets; decals live on one shared 1K–2K page; palettes use a small LUT with two 512–1K masks. If you exceed one envelope, plan a compensating reduction elsewhere and communicate the trade in your readme.

Common pitfalls and inclusive fixes

Expensive ornament everywhere, thin typographic decals, and hue‑only class encoding are the usual culprits. Replace micro filigree with embossed shapes that survive at distance. Convert serif slogans to blocky emblems or all‑caps sans with an outline buffer. Shift class language into shape, value, and motion so colorways don’t gate comprehension. Avoid palette swaps that drift skin into extreme values or flatten metalness into cloth; keep material identity stable even as hues change.

Closing: economy of pages, generosity of reads

Optimization and accessibility are not opposites. By rationing materials, sharing atlases, and designing details that scale with distance, you gain performance headroom to spend on the moments that matter—faces that emote, silhouettes that communicate role, and decals that tell stories players of many abilities can read. Treat texture and material planning as part of your visual language, and every variant you ship will be lighter on memory and heavier on meaning.