Researchers have developed TongueReenact, a novel framework for transferring tongue dynamics in face reenactment, addressing a gap in current systems that often neglect tongue movements. The system utilizes a foundation-model-assisted pipeline to create a specialized tongue segmentation model without requiring curated annotations. It also incorporates a spatially constrained latent masked diffusion model for realistic tongue synthesis, featuring adaptive mask dilation for smooth transitions at the mouth boundary. Experiments show significant improvements over existing methods, with a vision-language model-based evaluation protocol confirming its perceptual superiority. AI
IMPACT Enhances realism in face reenactment by addressing tongue dynamics, potentially improving applications in virtual avatars and digital communication.
RANK_REASON The cluster contains a research paper detailing a new technical framework for a specific AI task. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
