Silence Is Golden: Leveraging Adversarial Examples to Nullify Audio Control in LDM-Based Talking-Head Generation

Advances in talking-head animation based on Latent Diffusion Models LDM enable the creation of highly realistic, synchronized videos. These fabricated videos are indistinguishable from real ones, increasing the risk of potential misuse for scams, political manipulation, and misinformation. Hence,...