The convergence of deepfakes, Large Language Models (LLMs), and Retrieval-Augmented Generation (RAG) is revolutionizing medical education by offering new ways to deliver personalized, interactive, and adaptive learning experiences. Traditional methods are often non-customizable and unengaging, necessitating the need for innovative approaches that facilitate realistic simulation and dynamic feedback. Although deepfakes are a concern as they can generate artificial content, their use in education when done ethically has enormous potential. This paper explores the convergence the new AI tools: deepfakes for creating highly realistic virtual patients and clinical vignettes; LLMs with adaptive real-time feedback and explanation relevant to the context, and RAG, which builds upon updated clinical information based on trusted databases. We conduct a systematic SLR to contrast their applications, strengths, and weaknesses in medicine. Our findings show that incorporation of such technologies turns learning into an active and not a passive process, thus promoting critical thinking as well as clinical decision-making. However, the use of such technologies raises paramount ethical concerns related to content validity, patient data privacy, and risk of misinformation. Our feeling is that a pedagogically robust, well-controlled application of generative AI can be instrumental in advancing medical education while minimizing such risks. This study contributes to an awareness of the transformative potential and challenges involved in deploying deepfakes, LLMs, and RAG to the healthcare educational system, leading to more effective and ethical pedagogical practices.