VASA-3D: Lifelike Audio-Driven Gaussian Head Avatars from a Single Image | Read Paper on Bytez