Gather high-quality text-to-image prompts and test the performance of different models with the same prompts