OPT2I is a T2I optimization framework that leverages large language models (LLMs) to improve prompt-image consistency. By iteratively generating revised prompts, it refines the generation process. This significantly boosts consistency scores while maintaining FID and increasing the recall rate of generated data against real data.