The AIGC team of Xiaohongshu proposed a new algorithm for generating controllable faces in images and videos.

On August 18, according to information from Xiaohongshu, the Xiaohongshu AIGC team released a controllable face generation method called DynamicFace. It is reported that DynamicFace has achieved high-quality and highly consistent replacement effects for face fusion tasks in the image and video fields. Compared to traditional face replacement methods, DynamicFace innovatively combines diffusion models with composable 3D face priors to achieve a more consistent generation of face images and videos by finely decoupling facial motion and identity information.