FAM Diffusion: Frequency and Attention Modulation for High-Resolution Image Generation with Stable Diffusion

TL;DR

Fam Diffusion introduces frequency and attention modulation modules that enable high-resolution image generation with existing diffusion models without retraining, significantly reducing artifacts and latency.

Contribution

The paper presents a novel, training-free approach combining frequency and attention modulation to improve high-resolution image generation in diffusion models.

Findings

Addresses structural and local artifacts effectively

Achieves state-of-the-art quantitative performance

Maintains low latency without additional inference tricks

Abstract

Diffusion models are proficient at generating high-quality images. They are however effective only when operating at the resolution used during training. Inference at a scaled resolution leads to repetitive patterns and structural distortions. Retraining at higher resolutions quickly becomes prohibitive. Thus, methods enabling pre-existing diffusion models to operate at flexible test-time resolutions are highly desirable. Previous works suffer from frequent artifacts and often introduce large latency overheads. We propose two simple modules that combine to solve these issues. We introduce a Frequency Modulation (FM) module that leverages the Fourier domain to improve the global structure consistency, and an Attention Modulation (AM) module which improves the consistency of local texture patterns, a problem largely ignored in prior works. Our method, coined Fam diffusion, can seamlessly integrate into any latent diffusion model and requires no additional training. Extensive qualitative results highlight the effectiveness of our method in addressing structural and local artifacts, while quantitative results show state-of-the-art performance. Also, our method avoids redundant inference tricks for improved consistency such as patch-based or progressive generation, leading to negligible latency overheads.