arXiv:2511.18706v2 Announce Type: replace 
Abstract: Existing diffusion codecs typically build on text-to-image diffusion foundation models like Stable Diffusion. However, text conditioning is suboptimal from a compression perspective, hindering the potential of downstream diffusion codecs, particularly at ultra-low bitrates. To address it, we introduce \textbf{CoD}, the first \textbf{Co}mpression-oriented \textbf{D}iffusion foundation model, trained from scratch to enable end-to-end optimization of both compression and generation. CoD is not a fixed codec but a general foundation model designed for various diffusion-based codecs. It offers several advantages: \textbf{High compression efficiency}, replacing Stable Diffusion with CoD in downstream codecs like DiffC achieves SOTA results, especially at ultra-low bitrates (e.g., 0.0039 bpp); \textbf{Low-cost and reproducible training}, 300$\times$ faster training than Stable Diffusion ($\sim$ 20 vs. $\sim$ 6,250 A100 GPU days) on entirely open image-only datasets; \textbf{Providing new insights}, e.g., We find pixel-space diffusion can achieve VTM-level PSNR with high perceptual quality and can outperform GAN-based codecs using fewer parameters. We hope CoD lays the foundation for future diffusion codec research. Codes will be released.

يمثل تقديم نموذج CoD، وهو نموذج أساسي للانتشار موجه نحو الضغط، تقدمًا كبيرًا في تكنولوجيا ضغط الصور. على عكس النماذج الحالية التي تعتمد على التكييف النصي، تم تصميم CoD لتحقيق تحسين شامل لكل من الضغط والتوليد، محققًا نتائج رائدة في معدلات بت منخفضة للغاية.

La introducción de CoD, un nuevo modelo de difusión orientado a la compresión, representa un avance significativo en la tecnología de compresión de imágenes. A diferencia de los modelos existentes que dependen del condicionamiento textual, CoD está diseñado para la optimización de extremo a extremo tanto de la compresión como de la generación, logrando resultados de vanguardia en tasas de bits ultra-bajas.

L'introduction de CoD, un nouveau modèle de fondation de diffusion orienté vers la compression, marque une avancée significative dans la technologie de compression d'images. Contrairement aux modèles existants qui s'appuient sur le conditionnement textuel, CoD est conçu pour une optimisation de bout en bout à la fois de la compression et de la génération, atteignant des résultats à la pointe de la technologie à des débits ultra-bas.

The introduction of CoD, a novel compression-oriented diffusion foundation model, marks a significant advancement in image compression technology. Unlike existing models that rely on text conditioning, CoD is designed for end-to-end optimization of both compression and generation, achieving state-of-the-art results at ultra-low bitrates.

CoD: A Diffusion Foundation Model for Image Compression

Was this article worth reading? Share it

LucidQuery AI

ComfyUI

CompressX

Unifab

CreativePixel

Pixel Dojo

Ready to build your own newsroom?