LucidFlux: Caption-Free Universal Image Restoration via a Large-Scale Diffusion Transformer

Song Fei*, Tian Ye*, Lujia Wang, Lei Zhu

ICLR 2026

A caption-free 14B restoration model delivering photorealistic results beyond leading commercial systems.