Inadequate lighting conditions, poor quality, and artifacts often hinder accurately detecting objects, individuals, and activities, posing significant challenges in scenarios requiring clear and precise visual details. To mitigate these issues, video upscaling techniques are employed. In this context, Generative Adversarial Networks (GANs) are used to enhance the quality of surveillance video. Specifically, an improvised version of Enhanced Super-Resolution GAN (ESRGAN) is implemented. The refined model surpasses the original ESRGAN in performance. Its effectiveness is quantitatively assessed using the Peak Signal Noise Ratio (PSNR) and Structural Similarity Index (SSIM), demonstrating its superiority over existing methods.