 Text Control GAN is a novel GAN-based model designed for text-to-image synthesis tasks. It employs a neural network structure, known as a regressor, to learn features from conditional texts. Data augmentation techniques are used to improve the learning performance of the regressor, resulting in better image generation. Additionally, the discriminator focuses solely on GAN training, which leads to higher quality images with fewer artifacts. The model outperforms other models based on conditional GANs, such as GANINT-CLS, demonstrating a 17.6% increase in inception score and a 36.6% decrease in freshet inception distance. This article was authored by Ha Yoon-Koo and Min Hyuk Lee.