Brief

Researchers have introduced SeamEdit, a novel pipeline designed for semantic editing of large images using Visual-Language Models (VLMs). This training-free, model-agnostic approach treats VLMs as black-box oracles, addressing issues like semantic deformation and visible seams that arise when applying closed-source models to tiled editing. SeamEdit employs a five-stage process, including tile decomposition, VLM inpainting, consistency correction, candidate ranking, and seam fusion, to achieve high-quality edits with natural integration into the surrounding content. AI