报告嘉宾：徐晓刚 (The Chinese University of Hong Kong)
报告人简介：
Xiaogang Xu is a Research Fellow at the Chinese University of Hong Kong (CUHK), having previously held distinguished positions as a Research Scientist at Huawei and Zhejiang Lab, and as a ZJU Young 100 Professor at Zhejiang University. He earned his Ph.D. from CUHK and his Bachelor’s degree from Zhejiang University. His research focuses on the frontiers of Computer Vision, Large Models and Agents, Multi-modality Models, and AIGC, with a long-term vision directed toward Artificial General Intelligence (AGI). As a highly active member of the research community, he has served as an Area Chair for ICML and ACL and is a frequent reviewer for top-tier journals and conferences. His contributions have been recognized with the CVPR 2024 Best Demo Honorable Mention, a nomination for the World Artificial Intelligence Conference (WAIC) Youth Outstanding Paper Award, and inclusion in the 2025 Stanford-Elsevier World’s Top 2% Scientists list.
个人主页：
http://www.xuxiaogang.com
报告摘要：
Video generation models have advanced rapidly in recent years, with Seedance 2.0 serving as a prime example that text-to-video and image-to-video models have now reached practical standards. However, bridging the gap between these foundational models and real-world applications remains a significant challenge, as practical scenarios often demand specialized functional adaptation and enhanced controllability. In this talk, I will share several of our recent breakthroughs in applying video models to real-world tasks, including video motion control, slow-motion video interpolation, long-form video generation, and human-centric generation. These case studies demonstrate how video generation models are driving unprecedented transformation and performance gains in media applications and content creation. Ultimately, these advancements inspire us to further explore the vast potential for innovative application outputs in this evolving field.
