Towards Semantic 3D Hand-Object Interaction Generation via Functional Text Guidance