SPICE: A Synergistic, Precise, Iterative, and Customizable Image Editing Workflow
Kenan Tang, Yanhong Li *, Yao Qin · 2025
Read SPICE: A Synergistic, Precise, Iterative, and Customizable Image Editing Workflow by Kenan Tang, Yanhong Li *, Yao Qin (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-2504-09697.txt.
Description
Kenan Tang, Yanhong Li, and Yao Qin present an image editing workflow called SPICE in this 2025 paper. Prompt based editors follow broad instructions well, but they stumble on detailed local edits, and image quality often drops after a single change.
SPICE needs no extra training and accepts any image size or shape. It pairs a base diffusion model with an edge guided control model, letting the strengths of each cover the other. The workflow follows free form instructions closely, keeps untouched regions intact, and can improve quality across more than one hundred editing steps. On a hard real world editing dataset it beat leading baselines on measured scores and was preferred by human judges. The code is released for common editing interfaces.
Catalog record
| CREATOR | Kenan Tang, Yanhong Li *, Yao Qin |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 95 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-09697