MultiDiffusion: Fusing Diffusion Paths for Controlled Image Generation

TL;DR

MultiDiffusion introduces a unified, training-free framework that combines multiple diffusion processes through an optimization approach, enabling controllable and diverse image generation aligned with user constraints.

Contribution

It proposes a novel optimization-based method to fuse diffusion paths for controllable image synthesis without additional training or fine-tuning.

Findings

Enables high-quality, diverse images adhering to user controls

Supports various spatial constraints like masks and bounding boxes

Operates without further training or fine-tuning

Abstract

Recent advances in text-to-image generation with diffusion models present transformative capabilities in image quality. However, user controllability of the generated image, and fast adaptation to new tasks still remains an open challenge, currently mostly addressed by costly and long re-training and fine-tuning or ad-hoc adaptations to specific image generation tasks. In this work, we present MultiDiffusion, a unified framework that enables versatile and controllable image generation, using a pre-trained text-to-image diffusion model, without any further training or finetuning. At the center of our approach is a new generation process, based on an optimization task that binds together multiple diffusion generation processes with a shared set of parameters or constraints. We show that MultiDiffusion can be readily applied to generate high quality and diverse images that adhere to user-provided controls, such as desired aspect ratio (e.g., panorama), and spatial guiding signals, ranging from tight segmentation masks to bounding boxes. Project webpage: https://multidiffusion.github.io