AlignVid: Training-Free Attention Scaling for Semantic Fidelity in Text-Guided Image-to-Video Generation
Yexin Liu 1 , Wenjie Shu 1 , Zile Huang 2 , Haoze Zheng 1 , Yueze Wang 3 Manyuan Zhang 4 , Sernam Lim 2 , Harry Yang 1, 1 Hong Kong University of Science and Technology 2 University of Central Florida 3 Beijing Academy of Artificial Intelligence 4 The Chinese University of Hong Kong yliu292@connect. · 2025
Read AlignVid: Training-Free Attention Scaling for Semantic Fidelity in Text-Guided Image-to-Video Generation by Yexin Liu 1 , Wenjie Shu 1 , Zile Huang 2 , Haoze Zheng 1 , Yueze Wang 3 Manyuan Zhang 4 , Sernam Lim 2 , Harry Yang 1, 1 Hong Kong University of Science and Technology 2 University of Central Florida 3 Beijing Academy of Artificial Intelligence 4 The Chinese University of Hong Kong yliu292@connect. (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-01334.txt.
Description
Yexin Liu and co-authors fix a flaw in text-guided video made from a still image. New tools can start from one image and a text prompt. They keep the subject steady and the motion smooth. But they often miss fine points in the prompt.
When a prompt asks to add, drop, or change an object, the tool may keep the old scene. The team calls that miss semantic negligence. Their fix, AlignVid, scales attention with no new training. It steers the tool to heed the prompt more. That helps the video match what the user asked for.
Catalog record
| CREATOR | Yexin Liu 1 , Wenjie Shu 1 , Zile Huang 2 , Haoze Zheng 1 , Yueze Wang 3 Manyuan Zhang 4 , Sernam Lim 2 , Harry Yang 1, 1 Hong Kong University of Science and Technology 2 University of Central Florida 3 Beijing Academy of Artificial Intelligence 4 The Chinese University of Hong Kong yliu292@connect. |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 80 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-01334