SeedEdit: Align Image Re-Generation to Image Editing

TL;DR

SeedEdit is a diffusion-based image editing model that balances image preservation and new content generation, enabling diverse, stable, and sequential image revisions guided by text prompts.

Contribution

We propose SeedEdit, a novel diffusion model that effectively aligns image reconstruction and re-generation for improved image editing capabilities.

Findings

Achieves more diverse image edits than prior methods.

Provides stable and sequential image revision.

Balances original image preservation with new content generation.

Abstract

We introduce SeedEdit, a diffusion model that is able to revise a given image with any text prompt. In our perspective, the key to such a task is to obtain an optimal balance between maintaining the original image, i.e. image reconstruction, and generating a new image, i.e. image re-generation. To this end, we start from a weak generator (text-to-image model) that creates diverse pairs between such two directions and gradually align it into a strong image editor that well balances between the two tasks. SeedEdit can achieve more diverse and stable editing capability over prior image editing methods, enabling sequential revision over images generated by diffusion models.