Text-to-3D Conjuring

Turn text prompts into 3D models using AI generation APIs.

How It Works

1. Open the Conjure tab in the Wizard Console
2. Type a prompt (e.g., "medieval treasure chest", "sci-fi hover bike")
3. Select a provider and quality tier
4. Hit Generate
5. The model appears in your conjured library when ready
6. Click to place it in your world

Providers

Provider	Format	Speed	Quality
Meshy	GLB	~30-60s	High detail, textured
Tripo	GLB	~20-40s	Fast, good topology

Both providers return GLB files that are cached locally in public/conjured/.

Generation Pipeline

Each conjured asset goes through 4 phases:

Queued → Generating → Downloading → Ready

1. Queued — Request sent to provider
2. Generating — Provider is working (polled every 5 seconds)
3. Downloading — GLB downloaded and saved locally
4. Ready — Asset available for placement

Pipeline Extensions

Conjured assets can optionally go through additional processing:

• Auto-Rig — Automatically add a skeleton for animation
• Auto-Animate — Generate walk cycles and custom animations
• Thumbnail — CDN thumbnails are downloaded and saved locally (they expire after 3 days on provider CDNs)

Asset Storage

What	Where
GLB files	public/conjured/{assetId}.glb
Metadata	data/conjured-registry.json
Thumbnails	Saved alongside GLBs

Configuration

Requires API keys in your .env:

MESHY_API_KEY=your_meshy_key
TRIPO_API_KEY=your_tripo_key

Get keys from Meshy and Tripo.