From Text to 3D Model in 30 Seconds
Text-to-3D was one of the earliest AI 3D demos. The 2024 generation was rough — sometimes 90 seconds, sometimes 5 minutes, sometimes a melted-looking mess. The 2026 generation is fast and good.
Yugma uses Meshy as the text-to-3D backend, integrated directly into the AI Director's vocabulary. You don't leave the editor.
# How it works
When the AI Director needs a real-world object that isn't in the Sketchfab library — a custom-branded item, an unusual prop, a stylized variant — it calls Meshy's API via Yugma's generate_asset tool. The result lands in the scene as a normal scene-graph object.
Typical timing:
- 30-45 seconds for the first preview mesh.
- 60-90 seconds for the final textured model.
# When the AI uses it
The AI Director's decision tree (simplified):
- Does the user want a real-world object? Yes → continue.
- Is it findable in Sketchfab with reasonable quality? Yes → use Sketchfab (free, instant).
- No → call Meshy via
generate_asset.
This makes the AI cost-aware. Sketchfab is free; Meshy generations are credit-metered. The AI prefers Sketchfab.
# Free tier vs Pro
- Free: 5 AI compositions/day total. Generated assets count.
- Pro: Unlimited compositions; priority generation queue.
# Quality
Meshy ships a few model variants:
- Fast (lower quality, ~30s).
- Quality (default, ~60s).
- Pro (Pro Yugma tier, ~90s, higher fidelity).
For most scene work, default is right. For hero assets, prefer Pro tier or generate in Meshy directly with their full settings.
# What it can't do
- Photorealistic faces — uncanny valley territory.
- Articulated mechanisms — joints, hinges, sliders.
- CAD-grade precision — meter-scale, not millimeter.
- Specific copyrighted items — Meshy filters certain trademarks.
# When to use Sketchfab instead
Most everyday objects (chairs, tables, lamps, plants, cars, animals) already exist in Sketchfab, often free under CC. Faster, no credit cost, often higher quality. The AI prefers Sketchfab automatically.
# When to use Meshy directly
If you need batch asset generation (50+ unique items for a game), use the Meshy app directly with their full controls — Yugma's generate_asset is for one-at-a-time scene composition, not asset farming.
# What's next
- Multi-model bridging: route to Meshy / Tripo / Rodin based on quality/cost preferences.
- Image-to-3D in the studio: drop an image, get a 3D model in your scene.
- Style consistency: when you generate 5 props for the same scene, make them feel like they belong.