This project explores the extent to which ingress bandwidth over the internet can be saved by using prompt-based image generation and large language models to reconstruct source content at an edge server, while aiming to preserve the fidelity and quality of the transmitted content.