Saved in:
Bibliographic Details
Main Authors: Jin, Shutong, Wang, Ruiyu, Pokorny, Florian T.
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2312.12635
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916591110717440
author Jin, Shutong
Wang, Ruiyu
Pokorny, Florian T.
author_facet Jin, Shutong
Wang, Ruiyu
Pokorny, Florian T.
contents Even though large-scale text-to-image generative models show promising performance in synthesizing high-quality images, applying these models directly to image editing remains a significant challenge. This challenge is further amplified in video editing due to the additional dimension of time. This is especially the case for editing real-world videos as it necessitates maintaining a stable structural layout across frames while executing localized edits without disrupting the existing content. In this paper, we propose RealCraft, an attention-control-based method for zero-shot real-world video editing. By swapping cross-attention for new feature injection and relaxing spatial-temporal attention of the editing object, we achieve localized shape-wise edit along with enhanced temporal consistency. Our model directly uses Stable Diffusion and operates without the need for additional information. We showcase the proposed zero-shot attention-control-based method across a range of videos, demonstrating shape-wise, time-consistent and parameter-free editing in videos of up to 64 frames.
format Preprint
id arxiv_https___arxiv_org_abs_2312_12635
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle RealCraft: Attention Control as A Tool for Zero-Shot Consistent Video Editing
Jin, Shutong
Wang, Ruiyu
Pokorny, Florian T.
Computer Vision and Pattern Recognition
Even though large-scale text-to-image generative models show promising performance in synthesizing high-quality images, applying these models directly to image editing remains a significant challenge. This challenge is further amplified in video editing due to the additional dimension of time. This is especially the case for editing real-world videos as it necessitates maintaining a stable structural layout across frames while executing localized edits without disrupting the existing content. In this paper, we propose RealCraft, an attention-control-based method for zero-shot real-world video editing. By swapping cross-attention for new feature injection and relaxing spatial-temporal attention of the editing object, we achieve localized shape-wise edit along with enhanced temporal consistency. Our model directly uses Stable Diffusion and operates without the need for additional information. We showcase the proposed zero-shot attention-control-based method across a range of videos, demonstrating shape-wise, time-consistent and parameter-free editing in videos of up to 64 frames.
title RealCraft: Attention Control as A Tool for Zero-Shot Consistent Video Editing
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2312.12635