Real-time automatic 3D scene generation from natural language voice and text descriptions