---

name: asset-pipeline-3d description: Optimize and prepare 3D assets for web delivery. Use this skill when working with GLTF/GLB files, compressing 3D models, optimizing textures, setting up Blender exports, or preparing assets for Three.js/R3F. Covers GLTF workflows, Draco/meshopt compression, texture optimization, LOD generation, and performance profiling.

3D Asset Pipeline

Prepare and optimize 3D assets for performant web delivery.

Tool Versions (2026)

• gltf-transform: v4+
• Meshoptimizer: v0.21+
• Basis Universal: v1.16+
• Blender: 4.2+

Decision Frameworks

When to Use Which Compression

Geometry Compression:
├─ Need fastest decode time? → Meshopt (recommended)
├─ Need smallest file size? → Draco (10-20% smaller)
├─ Animated models (skinned/morphs)? → Meshopt (better animation support)
├─ Static meshes only? → Either works
└─ Maximum compatibility? → No compression (larger files)

Texture Compression:
├─ GPU-compressed (fastest rendering)? → KTX2/Basis Universal
├─ Good balance (size vs quality)? → WebP
├─ Best quality at cost of size? → PNG (lossless) or AVIF
├─ Need alpha channel? → WebP, PNG, KTX2, or AVIF
└─ Photo textures, no alpha? → JPEG or WebP

When to Use Which Pipeline

Single hero model (product viewer)?
  → gltf-transform optimize with high quality settings

Many similar models (e-commerce)?
  → Batch processing script with consistent settings

Game assets (performance critical)?
  → Aggressive compression + LODs + texture atlases

Architectural visualization?
  → High poly + baked lighting + KTX2 textures

Real-time application?
  → Meshopt + KTX2 + LODs + strict budgets

Compression Decision Matrix

Scenario	Geometry	Textures	LODs
Product viewer	Meshopt	WebP/KTX2	Optional
Game assets	Meshopt	KTX2	Required
Archviz	Meshopt	KTX2 2048px	Per-room
Mobile	Meshopt	KTX2 512-1024px	Required
E-commerce batch	Draco (smaller)	WebP	Optional

GLTF/GLB Overview

GLTF (GL Transmission Format) is the standard for web 3D. GLB is the binary version (single file).

GLTF supports:

• Meshes, materials (PBR), textures
• Skeletal animations, morph targets
• Scene hierarchy, cameras, lights
• Extensions for advanced features

File structure:

• .gltf - JSON with external binary/textures
• .glb - Single binary file (preferred for web)

Compression Tools

gltf-transform (Recommended)

Node.js CLI for GLTF optimization. Install: npm install -g @gltf-transform/cli

# Meshopt compression (recommended - faster decode)
gltf-transform meshopt input.glb output.glb

# Draco compression (smaller files, slower decode)
gltf-transform draco input.glb output.glb

# Texture compression to KTX2/Basis (GPU-compressed)
gltf-transform ktx input.glb output.glb --compress uastc

# Full optimization pipeline (2026 recommended)
gltf-transform optimize input.glb output.glb \
  --compress meshopt \
  --texture-compress ktx2

# WebP textures (good balance, wider support)
gltf-transform optimize input.glb output.glb \
  --compress meshopt \
  --texture-compress webp

# Resize textures (power of 2)
gltf-transform resize input.glb output.glb --width 1024 --height 1024

# Flatten scene hierarchy (reduces draw calls)
gltf-transform flatten input.glb output.glb

# Merge meshes sharing materials
gltf-transform join input.glb output.glb

# Remove unused data
gltf-transform prune input.glb output.glb

# Deduplicate accessors
gltf-transform dedup input.glb output.glb

# Generate simplified LOD
gltf-transform simplify input.glb output.glb --ratio 0.5 --error 0.001

# Full production pipeline
gltf-transform optimize input.glb output.glb \
  --compress meshopt \
  --texture-compress ktx2 \
  --texture-size 1024

gltfpack

Alternative CLI (part of meshoptimizer). Faster, less flexible.

# Basic optimization
gltfpack -i input.glb -o output.glb

# With Meshopt compression
gltfpack -i input.glb -o output.glb -cc

# With KTX2 texture compression
gltfpack -i input.glb -o output.glb -tc

# Simplify (50% triangle reduction)
gltfpack -i input.glb -o output.glb -si 0.5

# All optimizations (production ready)
gltfpack -i input.glb -o output.glb -cc -tc -si 0.5

Texture Optimization

Format Selection

Format	Use Case	Alpha	Compression	GPU Decode
KTX2/Basis	GPU-compressed (best)	Yes	UASTC/ETC1S	Native
WebP	General textures	Yes	Lossy/Lossless	CPU
AVIF	Best file compression	Yes	Lossy	CPU
JPEG	Photos, no alpha	No	Lossy	CPU
PNG	UI, sharp edges	Yes	Lossless	CPU

KTX2/Basis Universal is preferred for 2026 - textures stay compressed in GPU memory.

Size Guidelines

Texture Type	Mobile	Desktop	Notes
Diffuse/Albedo	512-1024	1024-2048	sRGB color space
Normal	512-1024	1024-2048	Linear, tangent space
Roughness/Metal	256-512	512-1024	Linear, can pack
AO	256-512	512-1024	Linear, can pack
Environment	256-512	512-1024	HDR for reflections

Power of 2: Always use power-of-2 dimensions (256, 512, 1024, 2048) for mipmapping.

Compression Commands

# ImageMagick - resize and convert
magick input.png -resize 1024x1024 -quality 85 output.webp

# cwebp (WebP)
cwebp -q 80 input.png -o output.webp

# Basis Universal (KTX2) - high quality UASTC
basisu -uastc -uastc_level 2 input.png -output_file output.ktx2

# Basis Universal - smaller ETC1S (lower quality)
basisu input.png -output_file output.ktx2

# toktx (Khronos tool)
toktx --t2 --uastc 2 output.ktx2 input.png

# Squoosh CLI (WebP/AVIF)
squoosh-cli --webp '{quality:80}' input.png
squoosh-cli --avif '{quality:60}' input.png

Channel Packing (ORM)

Combine grayscale maps into RGBA channels to reduce texture count:

R: Ambient Occlusion
G: Roughness
B: Metalness
A: (unused or height)

// Three.js with packed ORM texture
const ormTexture = textureLoader.load('/orm.png')
ormTexture.colorSpace = THREE.LinearSRGBColorSpace

const material = new THREE.MeshStandardNodeMaterial({
  aoMap: ormTexture,
  roughnessMap: ormTexture,
  metalnessMap: ormTexture,
  // Channel selection happens automatically in ORM workflow
})

Blender Export Settings

GLTF Export Checklist

1. Apply transforms: Ctrl+A → All Transforms
2. Apply modifiers: Check "Apply Modifiers" in export
3. Check scale: Blender default 1 unit = 1 meter
4. Clean up: Remove unused materials, objects, shape keys

Export Settings (Blender 4.2+)

Format: GLB (single file)

Include:
  [x] Selected Objects (if exporting subset)
  [x] Custom Properties
  [x] Cameras (if needed)
  [x] Punctual Lights (if needed)

Transform:
  +Y Up (standard for web)

Data:
  Mesh:
    [x] Apply Modifiers
    [x] UVs
    [x] Normals
    [ ] Tangents (let Three.js compute)
    [ ] Vertex Colors (unless needed)

  Material:
    [x] Materials → Export
    [x] Images → Automatic (embeds in GLB)

  Shape Keys: [x] (if using morph targets)
  Skinning: [x] (if using armatures)
  Armature:
    [x] Export Deformation Bones Only

Animation:
  [x] Animations
  Sampling Rate: 24 or match source

Compression:
  [ ] Draco (prefer post-processing with gltf-transform)

Common Export Issues

Problem	Solution
Model too big/small	Apply scale (Ctrl+A) before export
Missing textures	Pack textures (File → External Data → Pack)
Broken normals	Recalculate normals (Shift+N in Edit mode)
Flipped faces	Check normals direction, flip if needed
Animation jitter	Increase keyframe sampling rate
Materials wrong	Use Principled BSDF, check texture paths
Bones missing	Check "Export Deformation Bones Only"

Loading in Three.js

WebGPU Setup (2026)

import * as THREE from 'three/webgpu'
import { GLTFLoader } from 'three/addons/loaders/GLTFLoader.js'
import { MeshoptDecoder } from 'three/addons/libs/meshopt_decoder.module.js'
import { KTX2Loader } from 'three/addons/loaders/KTX2Loader.js'

const loader = new GLTFLoader()

// Meshopt decoder (recommended)
loader.setMeshoptDecoder(MeshoptDecoder)

// KTX2 textures (requires renderer for format detection)
const ktx2Loader = new KTX2Loader()
ktx2Loader.setTranscoderPath('https://cdn.jsdelivr.net/npm/three@0.171.0/examples/jsm/libs/basis/')
ktx2Loader.detectSupport(renderer) // Required!
loader.setKTX2Loader(ktx2Loader)

// Load
const gltf = await loader.loadAsync('/model.glb')
scene.add(gltf.scene)

Draco Fallback (if needed)

import { DRACOLoader } from 'three/addons/loaders/DRACOLoader.js'

const dracoLoader = new DRACOLoader()
dracoLoader.setDecoderPath('https://www.gstatic.com/draco/versioned/decoders/1.5.7/')
loader.setDRACOLoader(dracoLoader)

React Three Fiber

import { useGLTF } from '@react-three/drei'

// Drei handles Meshopt and Draco automatically
const { scene, nodes, materials, animations } = useGLTF('/model.glb')

// Preload for instant display
useGLTF.preload('/model.glb')

Performance Budgets

Target Metrics

Platform	Triangles	Draw Calls	Texture Memory	File Size
Mobile	<100K	<50	<50MB	<5MB
Desktop	<500K	<100	<200MB	<20MB
High-end	<2M	<200	<500MB	<50MB

Profiling

// Three.js render info (call after render)
console.log(renderer.info.render)
// { calls, triangles, points, lines, frame }

console.log(renderer.info.memory)
// { geometries, textures }

// Performance monitor
import Stats from 'stats.js'
const stats = new Stats()
stats.showPanel(0) // 0: fps, 1: ms, 2: mb
document.body.appendChild(stats.dom)

function animate() {
  stats.begin()
  renderer.render(scene, camera)
  stats.end()
}
renderer.setAnimationLoop(animate)

Asset Validation

# Validate GLTF structure
npx gltf-validator model.glb

# Online validation
# https://gltf.report

# Check file size breakdown
gltf-transform inspect model.glb

Related Skills

When you need...	Use skill
Build 3D scenes with Three.js	→ threejs
Build 3D scenes with React	→ react-three-fiber
Debug rendering/performance issues	→ graphics-troubleshooting

Reference Files

• references/compression.md - Detailed compression comparisons
• references/textures.md - Texture format deep dive
• references/validation.md - Asset validation and debugging
• references/lod.md - Level-of-detail strategies
• references/baking.md - Texture baking workflows