Use Cases Compare Learn Blog Docs Open Studio

From AI Prompt to GLB — A Full Walkthrough

The fastest way to understand the AI 3D pipeline is to walk a single scene from prompt to engine. Here's the full path in five minutes.

Step 1 — describe the scene

Open Yugma Studio. Type:

"Show a small modernist coffee shop interior — six round tables, eight chairs, a long bar with three pendant lamps overhead, espresso machine on the bar, exposed brick on the back wall, oak floor, warm afternoon lighting."

Hit enter. The AI Director places everything in 90 seconds.

Step 2 — refine

Click any object to inspect properties. The AI handles big moves; the inspector handles fine ones. Or keep using chat:

"Move the bar one meter to the left."
"Make the pendants brass instead of black."

Step 3 — export GLB

Cmd + K → Export → GLB. Yugma serializes the scene graph to glTF 2.0 binary.

Step 4 — drop into your engine

Unity (2022.x +): drag the GLB into /Assets/; it imports as a Prefab. Materials map to Standard/PBR.

Unreal: use glTFRuntime for runtime imports or Datasmith for editor-time. Lighting needs a recompute (Build Lighting → Production).

Godot 4: drag GLB into project. Imports as a PackedScene.

Blender: File → Import → glTF 2.0. PBR materials map to Principled BSDF.

Three.js: useGLTF('/path/to/scene.glb') in R3F, or new GLTFLoader().load(...) in vanilla.

Step 5 — extend in code

Once in your engine, add the things Yugma doesn't do today: physics, custom shaders, gameplay logic, audio. The GLB stays a clean asset; your engine code wraps it.

What can break

A real example

For an indie Godot game, this workflow takes a level blockout from sentence to playable in 4 minutes: 90 seconds in Yugma, 30 seconds export, 60 seconds in Godot to drop in player physics. That's ~80% of the value of a senior 3D artist's first day on a level.

Read about the AI 3D export pipeline →