BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
X-LIC-LOCATION:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:19911015T033000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20251218T030657Z
LOCATION:Meeting Room S423+S424\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251215T151100
DTEND;TZID=Asia/Hong_Kong:20251215T152200
UID:siggraphasia_SIGGRAPH Asia 2025_sess109_papers_1191@linklings.com
SUMMARY:CamCloneMaster: Enabling Reference-based Camera Control for Video 
 Generation
DESCRIPTION:Yawen Luo (Chinese University of Hong Kong, Kuaishou Technolog
 y); Xiaoyu Shi (Kuaishou Technology); Jianhong Bai (Zhejiang University, K
 uaishou Technology); Menghan Xia (Kuaishou Technology); Tianfan Xue (Chine
 se University of Hong Kong); and Xintao Wang, Pengfei Wan, Di Zhang, and K
 un Gai (Kuaishou Technology)\n\nCamera control is crucial for generating e
 xpressive and cinematic videos. Existing methods rely on explicit sequence
 s of camera parameters as control conditions, which can be cumbersome for 
 users to construct, particularly for intricate camera movements. To provid
 e a more intuitive camera control method, we propose CamCloneMaster, a fra
 mework that enables users to replicate camera movements from reference vid
 eos without requiring camera parameters or test-time fine-tuning. CamClone
 Master seamlessly supports reference-based camera control for both Image-t
 o-Video and Video-to-Video tasks within a unified framework. Furthermore, 
 we present the Camera Clone Dataset, a large-scale synthetic dataset desig
 ned for camera clone learning, encompassing diverse scenes, subjects, and 
 camera movements. Extensive experiments and user studies demonstrate that 
 CamCloneMaster outperforms existing methods in terms of both camera contro
 llability and visual quality.\n\nRegistration Category: Full Access, Full 
 Access Supporter\n\nSession Chair: Marc Christie (IRISA, INRIA, Univ Renne
 s, CNRS)\n\n
END:VEVENT
END:VCALENDAR