EventAVS
Aligning Audio and Video Generation Model with Event-level Synchrony
EventAVS: Aligning Audio and Video Generation Model with Event-level Synchrony
Status: Under Review
Affiliation: GeWu-Lab, Renmin University of China
Overview
EventAVS introduces a novel approach for aligning audio and video generation models with event-level synchrony, achieving more coherent and temporally consistent multimodal generation.
Key Contributions
- Developed event-level synchronization mechanisms for audio-visual generation
- Enhanced temporal consistency in multimodal generation systems
- Advanced audio-visual co-generation techniques
Authors
Youquan Fu, Song Huang, Ruiyang Si, Ping Luo, Jiacheng Sun, Hongyuan Zhang, Di Hu, Xuelong Li
Research Context
This work contributes to tasks in audio-visual co-generation and complex audio editing based on Multimodal Large Language Models (MLLMs), conducted during my research internship at GeWu-Lab.
Supervision
Supervised by Prof. Di Hu at the GeWu Lab, Renmin University of China (February 2025 - September 2025).
This project aligns with my research interests in unified models and multimodal AI systems.