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:20251218T030655Z
LOCATION:Meeting Room S423+S424\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251218T094300
DTEND;TZID=Asia/Hong_Kong:20251218T095400
UID:siggraphasia_SIGGRAPH Asia 2025_sess147_papers_2210@linklings.com
SUMMARY:Zero-Shot Dynamic Concept Personalization with Grid-Based LoRA
DESCRIPTION:Rameen Abdal, Or Patashnik, Ekaterina Deyneka, Hao Chen, Aliak
 sandr Siarohin, Sergey Tulyakov, Daniel Cohen-Or, and Kfir Aberman (Snap I
 nc.)\n\nRecent advances in text-to-video generation have enabled high-qual
 ity synthesis from text and image prompts. While the personalization of dy
 namic concepts, which capture subject-specific appearance and motion from 
 a single video, is now feasible, most existing methods require per-instanc
 e fine-tuning, limiting scalability. We introduce a fully zero-shot framew
 ork for dynamic concept personalization in text-to-video models. Our metho
 d leverages structured 2×2 video grids that spatially organize input and o
 utput pairs, enabling the training of lightweight Grid-LoRA adapters for e
 diting and composition within these grids. At inference, a dedicated Grid 
 Fill module completes partially observed layouts, producing temporally coh
 erent and identity preserving outputs. Once trained, the entire system ope
 rates in a single forward pass, generalizing to previously unseen dynamic 
 concepts without any test-time optimization. Extensive experiments demonst
 rate high-quality and consistent results across a wide range of subjects b
 eyond trained concepts and editing scenarios.\n\nRegistration Category: Fu
 ll Access, Full Access Supporter\n\nSession Chair: Fan Tang (Institute of 
 Computing Technology, Chinese Academy of Sciences)\n\n
END:VEVENT
END:VCALENDAR