“In the field of 3D speech face generation, experts have proposed a layered decoupling method based on VQ-VAE, which effectively improves the accuracy of face reconstruction and the stability of expression generation.”