find-3d-assets

Finding 3D Assets for Web Projects

Search Strategy

When looking for a specific 3D model, follow this priority order:

1. Check Local Models First

Look in modules/web-app/public/models/ — the model (or something close) might already exist. See the Current Project Models table at the bottom.

2. Search Sketchfab (Largest Library, API Download)

Sketchfab is the primary source for high-quality downloadable 3D models. It has API-based download which makes it the most automatable option when the user has a SKETCHFAB_API_TOKEN.

Finding models — web search:

site:sketchfab.com [object name] low poly downloadable free
sketchfab [object name] CC0 OR CC-BY download

Finding models — direct Sketchfab search URL:

https://sketchfab.com/search?q=[keyword]&downloadable=true&sort_by=-likeCount

Finding models — Sketchfab Search API:

# Search without auth (public models)
curl -s "https://api.sketchfab.com/v3/search?type=models&q=standing+desk&downloadable=true&sort_by=-likeCount&count=5" \
  | python3 -c "
import sys, json
data = json.load(sys.stdin)
for r in data.get('results', []):
    print(f\"{r['name']} by {r['user']['displayName']} — {r['faceCount']} tris — {r['license']['label']}\"
          f\"  https://sketchfab.com/3d-models/{r['slug']}-{r['uid']}\")
"

Filters to use when searching:

• downloadable=true — only models with download enabled
• sort_by=-likeCount — most popular first (usually highest quality)
• license=by for CC-BY, license=cc0 for CC0
• max_face_count=20000 — keep it web-friendly

Sketchfab Download Workflow (Automated)

Requires: SKETCHFAB_API_TOKEN environment variable (stored in .env, loaded by direnv). Get one at: https://sketchfab.com/settings/password → "API Token" section.

MODEL_ID="65a7f4b06a5f4954a0d43eb8812dd165"   # hex ID from the model URL
OUTPUT="standingDesk.glb"                       # target filename

# Step 1 — Get signed download URLs (expires in 5 min)
DOWNLOAD_JSON=$(curl -s \
  -H "Authorization: Token $SKETCHFAB_API_TOKEN" \
  "https://api.sketchfab.com/v3/models/$MODEL_ID/download")

# Step 2 — Extract the GLB URL (preferred) or fall back to GLTF zip
GLB_URL=$(echo "$DOWNLOAD_JSON" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d.get('glb',{}).get('url',''))")
GLTF_URL=$(echo "$DOWNLOAD_JSON" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d.get('gltf',{}).get('url',''))")

if [ -n "$GLB_URL" ]; then
  # Direct GLB — no conversion needed
  curl -L -o "/tmp/${OUTPUT}" "$GLB_URL"
else
  # GLTF zip — download, unzip, convert
  curl -L -o /tmp/model.zip "$GLTF_URL"
  unzip -o /tmp/model.zip -d /tmp/model_tmp/
  npx gltf-pipeline -i /tmp/model_tmp/scene.gltf -o "/tmp/${OUTPUT}"
  rm -rf /tmp/model.zip /tmp/model_tmp/
fi

# Step 3 — Optimize (MANDATORY — see "Post-Download Optimization" section)
# Sketchfab GLBs are typically 5-30MB; this pipeline gets them to 100-500KB.
npx gltf-transform resize "/tmp/${OUTPUT}" /tmp/opt-s1.glb --width 256 --height 256
npx gltf-transform simplify /tmp/opt-s1.glb /tmp/opt-s2.glb --ratio 0.05 --error 0.02
npx gltf-transform webp /tmp/opt-s2.glb /tmp/opt-s3.glb
npx gltf-transform prune /tmp/opt-s3.glb /tmp/opt-s4.glb
npx gltf-transform draco /tmp/opt-s4.glb "modules/web-app/public/models/${OUTPUT}"
rm -f /tmp/opt-s*.glb "/tmp/${OUTPUT}"

# Step 4 — Generate gzip sidecar for nginx gzip_static
node -e "
const fs=require('fs'),zlib=require('zlib');
const p='modules/web-app/public/models/${OUTPUT}';
const d=fs.readFileSync(p);
fs.writeFileSync(p+'.gz',zlib.gzipSync(d,{level:9}));
console.log('${OUTPUT}:',(d.length/1024).toFixed(0)+'KB →',(fs.statSync(p+'.gz').size/1024).toFixed(0)+'KB gzipped');
"