Towards Training-Free Scene Text Editing
Yubo Li, Xugong Qin, Peng Zhang, Hailun Lin, Gangyan Zeng, Kexin Zhang · 2026
Read Towards Training-Free Scene Text Editing by Yubo Li, Xugong Qin, Peng Zhang, Hailun Lin, Gangyan Zeng, Kexin Zhang (2026) 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-2603-24571.txt.
Description
Li and co-authors present TextFlow, a training-free framework for editing text in natural images. The goal is to change written content while keeping the scene realistic and consistent. Existing methods often need task-specific training or paired data. TextFlow pairs flow manifold steering, which preserves the style and structure of characters and background, with attention boost, which sharpens the rendering of the new text. Tests across diverse scenes and languages show visual quality and text accuracy matching or beating trained rivals, with no extra training.
Catalog record
| CREATOR | Yubo Li, Xugong Qin, Peng Zhang, Hailun Lin, Gangyan Zeng, Kexin Zhang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49888 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.24571