text-to-blender

text-to-blender — Orchestrator Skill

Entry point for natural language → Blender scene generation. Chain-loads sub-skills based on user intent.

Intent Detection

Parse user request and decide which sub-skills to load:

User: "create a red chair with three-point lighting and render"
→ modeling + materials + lighting + cameras + rendering

User: "animate the sphere moving from left to right"
→ animation

User: "export the scene as glTF"
→ export

User: "make a 3D model from this wireframe sketch"
→ wireframe-to-3d

Workflow Sequence (Pro Mode)

1. Block-out: Create primitive shapes with correct proportions
2. Camera: Set up framing and composition
3. Lighting: Add environment or artificial lights
4. Forms: Refine geometry with modifiers
5. Materials: Apply colors and PBR materials
6. Detail: Add edge loops, bevels, subdivisions
7. Render: Configure engine, resolution, output
8. Composite: Post-processing (optional)
9. Export: GLTF/FBX/USD/STL

Conventions

• Use execute_blender_code() for all geometry operations
• Name objects with prefixes: GEO_ (geometry), MAT_ (material), LGT_ (light), CAM_ (camera)
• Always apply transforms before exporting
• Validate with get_viewport_screenshot_image() after each phase

Example Chain

text-to-blender
  → blender-pro-workflow (sequencing)
  → blender-modeling
  → blender-materials
  → blender-lighting
  → blender-cameras
  → blender-rendering
  → blender-export