ELLA: Enhancing Text-to-Image Models with Large Language Models for Dense Prompt Alignment

ELLA equips diffusion models with Large Language Models to improve text alignment for dense prompts.