While micro-teaching is essential for honing instructional techniques, its effectiveness is hindered by limited practice opportunities and unrealistic peer role-playing, which fails to replicate authentic students’ cognition levels, knowledge mastery progression, behaviour, and feedback. To solve this problem, we first developed cutting-edge GenAI models and algorithms to simulate real student response in the classroom and generate pedagogical feedback. Secondly, a metaverse-based training system equipped with our proposed GenAI models and algorithms are developed, where preservice teachers engage in iterative cycles of responsive teaching practice, AI-informed reflection, and AI-enabled revision using VR technology. Lastly, we used this GenAI-empowered training system to build impact cases for preservice teacher training in Chinese reading and writing instruction in primary and secondary schools in Hong Kong and the Great Bay Areas and gradually scale up to other subjects and more teaching skills.