3D Portrait

3D portrait generation aims to create realistic, three-dimensional representations of human faces from various input sources, such as single images or videos, often incorporating text-based style guidance. Current research focuses on improving the accuracy and realism of these models, employing techniques like GANs, diffusion models, and novel 3D representations (e.g., Gaussian splatting, tri-plane representations) to address challenges in geometry reconstruction, texture generation, and temporal consistency in video applications. This field is significant for its potential applications in telepresence, AR/VR, and entertainment, offering advancements in both computer vision and graphics.