AvatarMMC presents a new method for generating and editing 3D head avatars. The approach uses a 3D Generative Adversarial Network (GAN) and a Latent Diffusion Model (LDM) for multi-modal conditioning. This provides better control over the generation and editing of synthetic avatars. The method can generate and edit avatars given a mixture of control signals such as RGB input, segmentation masks, and global attributes. Experiments show that this approach outperforms a solely GAN-based approach on generation and editing tasks.
Publication date: 9 Feb 2024
Project Page: https://arxiv.org/abs/2402.05803
Paper: https://arxiv.org/pdf/2402.05803
