Abstract: Rapidly synthesizing text-relevant images has long been a significant challenge. Introducing pre-trained models into GANs can significantly enhance model performance, enabling the fast ...