This project delves into the comparative analysis of the traditional GANs, Wasserstein GANs (WGANs), and Auxiliary Classifier GAN (AC-GAN) to unravel the nuances of their generative capabilities and performance. By harnessing the complexity of the CIFAR-10 dataset, we embark on a journey to not only construct and refine these models but also to critique and quantify their performance through a multi-faceted lens of evaluation metrics. Besides the former three variants of GANs, we seek beyond and explore WGAN with gradient penalty (WGAN-GP), Deep Regret Analytic GAN (DRAGAN), and EBGAN aiming at broadening the generative scope of our models and methods to improve the performance and training stability of GANs