This chapter explores the future research directions in AI-based image and video coding. It provides critical challenges including computational complexity, model interpretability, and model efficiency while highlighting the need for adaptive, perception-driven, and task-oriented compression methods that cater to specific applications, such as human vision, machine vision, and real-time scenarios. It discusses the integration of 3D coding frameworks, as well as challenges in reducing compression artifacts and achieving energy-efficient scalable models. Additionally, the chapter examines the evolution of coding standards and the role of open source projects in fostering innovations. By addressing these challenges and leveraging the potential of large-scale generative models, the chapter provides a roadmap for future advancements in image and video coding technologies, aiming to enhance efficiency, quality, and adaptability across diverse applications.