GANs for digital art in entertainment: techniques, applications, and aesthetic evaluation

Artificial Intelligence has revolutionized digital entertainment, such as generating highly realistic and creative visual content, and Generative Adversarial Networks (GANs) have proven to be a pivotal technology in the synthesis and generation of images for artistic purposes. The aim of this review is to conduct a thorough and systematic study of the GAN-based digital art systems, their architectures, applications and evaluation approaches, with the objective of understanding their role in the context of contemporary creative AI research. The study uses a structured review method (PRISMA), the literature from key scientific databases, which is clearly defined and selected by inclusion and exclusion criteria that make it relevant and quality. The results demonstrate various GAN architectures such as Vanilla GAN, DCGAN, Conditional GAN, CycleGAN, StyleGAN, attention-based, and multimodal GANs applied to domains including gaming, animation, film production, virtual reality, NFT art, and music visualization. The review also highlights the important trends in the aesthetic evaluation, both objective and subjective, human-centered approaches. Overall, the study identifies some research gaps such as inconsistency of evaluation, ethical issues, computational complexity, and lack of interpretation, and suggests future trends along three main directions: hybrid generative models, real-time systems and emotionally intelligent AI for next-generation digital entertainment.