VASA-3D: Lifelike Audio-Driven Gaussian Head
Avatars from a Single Image