AI RESEARCH

CoD-Lite: Real-Time Diffusion-Based Generative Image Compression

arXiv CS.CV

ArXi:2604.12525v1 Announce Type: new Recent advanced diffusion methods typically derive strong generative priors by scaling diffusion transformers. However, scaling fails to generalize when adapted for real-time compression scenarios that demand lightweight models. In this paper, we explore the design of real-time and lightweight diffusion codecs by addressing two pivotal questions. First, does diffusion pre-