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:20251218T030527Z
LOCATION:Meeting Room S423+S424\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251218T090000
DTEND;TZID=Asia/Hong_Kong:20251218T100500
UID:siggraphasia_SIGGRAPH Asia 2025_sess147@linklings.com
SUMMARY:Text-to-Image & Customization
DESCRIPTION:The Technical Papers program is the heartbeat of SIGGRAPH Asia
 , spotlighting world-class scholarly research at the forefront of computer
  graphics and interactive techniques. For decades, it has been the definit
 ive venue where bold ideas take root, foundational concepts are reimagined
 , and the future of visual computing is shaped.\n\nThis year, we explore n
 ew intersections of algorithms and artistry, automation and authorship, to
 ols and imagination – challenging the very way we design, simulate, visual
 ize, and interact with digital worlds.\n\nConsiStyle: Style Diversity in T
 raining-Free Consistent T2I Generation\n\nIn text-to-image models, consist
 ent character generation is the task of achieving text alignment while mai
 ntaining the subject's appearance across different prompts. However, since
  style and appearance are often entangled, the existing methods struggle t
 o preserve consistent subject characteristics ...\n\n\nYohai Mazuz (Tel Av
 iv University); Janna Bruner (Tel Aviv University, Amazon); and Lior Wolf 
 (Tel Aviv University)\n---------------------\nBokeh Diffusion: Defocus Blu
 r Control in Text-to-Image Diffusion Models\n\nRecent advances in large-sc
 ale text-to-image models have revolutionized creative fields by generating
  visually captivating outputs from textual prompts; however, while traditi
 onal photography offers precise control over camera settings to shape visu
 al aesthetics—such as depth-of-field via aper...\n\n\nArmando Fortes, Tian
 yi Wei, Shangchen Zhou, and Xingang Pan (Nanyang Technological University,
  Singapore)\n---------------------\nMagic Fixup: Streamlining Photo Editin
 g by Watching Dynamic Videos\n\nWe propose a generative model that, given 
 a coarsely edited image, synthesizes a photorealistic output that follows 
 the prescribed layout.  By using simple segmentations and coarse 2D manipu
 lations, we can synthesize a photorealistic edit faithful to the user's in
 put while addressing secondary effec...\n\n\nHadi Alzayer (University of M
 aryland College Park, Adobe Inc.); Zhihao Xia, Cecilia Zhang, and Eli Shec
 htman (Adobe Inc.); Jia-Bin Huang (University of Maryland College Park); a
 nd Michael Gharbi (Adobe Inc.)\n---------------------\nPractiLight: Practi
 cal Light Control Using Foundational Diffusion Models\n\nLight control in 
 generated images is a difficult task, posing specific challenges, spanning
  over the entire image and frequency spectrum. Most approaches tackle this
  problem by training on extensive yet domain-specific datasets, limiting t
 he inherent generalization and applicability of the foundatio...\n\n\nYota
 m Erel (Tel Aviv University), Rishabh Dabral and Vladislav Golyanik (Max P
 lanck Institute for Informatics), Amit H. Bermano (Tel Aviv University), a
 nd Christian Theobalt (Max Planck Institute for Informatics)\n------------
 ---------\nZero-Shot Dynamic Concept Personalization with Grid-Based LoRA\
 n\nRecent advances in text-to-video generation have enabled high-quality s
 ynthesis from text and image prompts. While the personalization of dynamic
  concepts, which capture subject-specific appearance and motion from a sin
 gle video, is now feasible, most existing methods require per-instance fin
 e-tunin...\n\n\nRameen Abdal, Or Patashnik, Ekaterina Deyneka, Hao Chen, A
 liaksandr Siarohin, Sergey Tulyakov, Daniel Cohen-Or, and Kfir Aberman (Sn
 ap Inc.)\n---------------------\nB4M: Breaking Low-Rank Adapter for Making
  Content-Style Customization\n\nThis paper proposes a novel framework for 
 personalized content-style fusion generation by training content and style
  in separated parameter space of low-rank adaptations for pre-trained text
 -to-image models. We introduce “partly learnable projection” (PLP) matrice
 s and a “break-for...\n\n\nYu Xu (Chinese Academy of Sciences Institute of
  Computing Technology); Fan Tang and Juan Cao (Institute of Computing Tech
 nology, Chinese Academy of Sciences); Yuxin Zhang (Institute of Automation
 , Chinese Academy of Sciences); Oliver Deussen (University of Konstanz); w
 eiming Dong (Institute of Automation, Chinese Academy of Sciences); Jintao
  Li (Institute of Computing Technology, Chinese Academy of Sciences); and 
 Tong-Yee Lee (National Cheng-Kung University)\n\nRegistration Category: Fu
 ll Access, Full Access Supporter\n\nSession Chair: Fan Tang (Institute of 
 Computing Technology, Chinese Academy of Sciences)
END:VEVENT
END:VCALENDAR