ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment

TL;DR

ELLA enhances diffusion models' understanding of complex, dense prompts by integrating large language models with a novel semantic connector, significantly improving prompt comprehension and image generation quality.

Contribution

This paper introduces ELLA, a method to incorporate LLMs into diffusion models for better semantic alignment without additional training, and proposes DPG-Bench for evaluating dense prompt following.

Findings

ELLA outperforms existing methods in dense prompt following.

Improves handling of multiple objects and complex relationships.

Demonstrates effectiveness on the DPG-Bench benchmark.

Abstract

Diffusion models have demonstrated remarkable performance in the domain of text-to-image generation. However, most widely used models still employ CLIP as their text encoder, which constrains their ability to comprehend dense prompts, encompassing multiple objects, detailed attributes, complex relationships, long-text alignment, etc. In this paper, we introduce an Efficient Large Language Model Adapter, termed ELLA, which equips text-to-image diffusion models with powerful Large Language Models (LLM) to enhance text alignment without training of either U-Net or LLM. To seamlessly bridge two pre-trained models, we investigate a range of semantic alignment connector designs and propose a novel module, the Timestep-Aware Semantic Connector (TSC), which dynamically extracts timestep-dependent conditions from LLM. Our approach adapts semantic features at different stages of the denoising process, assisting diffusion models in interpreting lengthy and intricate prompts over sampling timesteps. Additionally, ELLA can be readily incorporated with community models and tools to improve their prompt-following capabilities. To assess text-to-image models in dense prompt following, we introduce Dense Prompt Graph Benchmark (DPG-Bench), a challenging benchmark consisting of 1K dense prompts. Extensive experiments demonstrate the superiority of ELLA in dense prompt following compared to state-of-the-art methods, particularly in multiple object compositions involving diverse attributes and relationships.