Medical professions’ time is invaluable, especially in regions where public medical resources are scarce. They aim to convey substantial content in minimal time to assist a maximum number of patients. However, due to cognitive declines, personal experience, and insufficient medical knowledge, elderly patients often encounter challenges in swiftly comprehending the purpose for medical instructions, especially diagnostic plans. This miscommunication frequently results in feelings of reluctance, frustration, and mistrust towards proposed medical procedures, ultimately impacting the satisfaction of medical communication and execution of treatment plans adversely. In this research, we introduce an innovative approach where the task of explaining medical procedures is entrusted to short-form video generated with Large Language Models (LLMs). Based on preliminary data on patient preferences and medical history, LLM is applied in generating personalized explanatory video, featuring virtual representations of a doctor. The generation strictly follows a procedure of persona modeling, target generation, script generation, and video generation. With a N=20 study involving a simulated scenario where physicians suggest an MRI scan - a costly and sometimes unfamiliar procedure for many elderly patients - the proposed method show an obvious decline in the patient’s negative experience, including reluctance, frustration, and mistrust. This study pioneers the integration of Large Language Models in crafting personalized explanatory videos for elderly patients, enhancing medical comprehension and experience. Our method presents a novel convergence of Meta Human, NLP technology, and elderly’s patient-centered care, bridging communication gaps in healthcare scenarios.