log in  |  register  |  feedback?  |  help  |  web accessibility
PhD Proposal: PhD Preliminary Generative Models for Visual Content Creation: Video, Storytelling, and Documents
Taewon Kang
IRB-5105 https://umd.zoom.us/j/93523753435?pwd=7nhzsbeQFd4zUxlRpPKqbgI426aGOL.1
Monday, November 30, 2026, 4:30-6:00 pm
  • You are subscribed to this talk through .
  • You are watching this talk through .
  • You are subscribed to this talk. (unsubscribe, watch)
  • You are watching this talk. (unwatch, subscribe)
  • You are not subscribed to this talk. (watch, subscribe)
Abstract

Generative models now produce images and videos of high visual quality, yet the people who make content with them still struggle to control what comes out.

This proposal presents my research on that control problem across three media. In video, a story told through several prompts must stay coherent across scene boundaries. In storytelling, characters must speak in a way that matches what is on screen and what they said before. In documents, a new background must not damage the text, layout, and readability of a page. The completed work addresses each medium with a training-free method built on pretrained generative models. Text2Story keeps multi-scene video coherent, Action2Dialogue gives characters grounded dialogue and speech, and BGGEN generates document backgrounds that preserve the text, with an accepted follow-up, Dynamics-Inspired Diffusion, that removes the explicit masks.

The proposed research continues this line in two thrusts. The first brings three manuscripts now under review (NEGATE, DCR, and TPD), which control what pretrained video models get wrong by default, together into one control layer for general video generation, in the same direction as the planned industry internship. The second, part of an interdisciplinary research collaboration, turns news articles into short explainer videos for viewers with no prior knowledge of the events, with the goal that viewers understand the article from the video as well as readers do from the text. The goal for the dissertation defense is Spring 2028 or earlier.

Bio

Taewon Kang is a PhD student (Aug 2024-) in the Department of Computer Science at the University of Maryland, College Park. He received his M.S. in Computer Science from the University of Maryland, College Park in May 2026. Before that, he completed his undergraduate studies (Mar 2020-Feb 2024) in the Department of Computer Science and Engineering at Korea University in Seoul, South Korea. He graduated with Great Honors (magna cum laude) and received Semester Honors (4.5/4.5).

His research focuses on generative models and content creation. He has a strong background in diffusion models, text-to-video generation, storytelling, and document generation. His work has been published at ECCV, ACCV, ICIP, and NeurIPS (Spotlight, as a co-author). During his PhD, he has been supported by the Mogam Science Scholarship (2024, 2025). In 2026, he also received the Korean American Scholarship Foundation (KASF) Graduate Scholarship and the Korean Honor Scholarship from the Embassy of the Republic of Korea in the USA. He has served as a reviewer for top-tier venues, including NeurIPS, ICLR, CVPR, ICML, and ICCV, and as a Program Committee member for AAAI. He received the ICML 2026 Silver Reviewer Award.

He was awarded a National Presidential Science Scholarship by President Moon Jae-in in 2020. That same year, he received the 2020 Talent Award of Korea from the Korean Ministry of Education and the 2020 Youth Hero Prize in Academic (Science) from the Korea Scout Association. Before entering Korea University, he attended Sejong Academy of Science and Arts in Sejong, South Korea (Mar 2017–Jan 2020), where he was admitted through priority selection as an early entrant after graduating early from middle school.

This talk is organized by Migo Gui