Appendix Folder Description

This folder contains supplementary materials for our ICLR 2026 submission. The contents are organized as follows:


- compare_with_SOTA_method/  
  This folder contains video comparisons between our proposed AUHead model and state-of-the-art (SOTA) methods. All videos use the same audio and reference images, providing a direct visual comparison. These clips highlight AUHead’s superior temporal consistency and expression realism.

- Emotion/  
  This folder demonstrates emotional generation results. Using neutral speech audio along with AU features and reference images reflecting different emotions, these videos demonstrate AUHead’s ability to generate expressive talking-head videos. The results show strong emotional fidelity and stable identity preservation.