Skip to yearly menu bar Skip to main content


DemoFusion: Democratising High-Resolution Image Generation With No $$$

Ruoyi DU · Dongliang Chang · Timothy Hospedales · Yi-Zhe Song · Zhanyu Ma

Arch 4A-E Poster #132
[ ] [ Project Page ]
Wed 19 Jun 5 p.m. PDT — 6:30 p.m. PDT


High-resolution image generation with Generative Artificial Intelligence (GenAI) has immense potential but, due to the enormous capital investment required for training, it is increasingly centralised to a few large corporations, and hidden behind paywalls. This paper aims to democratise high-resolution GenAI by advancing the frontier of high-resolution generation while remaining accessible to a broad audience. We demonstrate that existing Latent Diffusion Models (LDMs) possess untapped potential for higher-resolution image generation. Our novel DemoFusion framework seamlessly extends open-source GenAI models, employing Progressive Upscaling, Skip Residual, and Dilated Sampling mechanisms to achieve higher-resolution image generation. The progressive nature of DemoFusion requires more passes, but the intermediate results can serve as "previews", facilitating rapid prompt iteration.

Live content is unavailable. Log in and register to view live content