Generative AI enables scalable production of educational videos, but current systems largely focus on producing visually coherent content rather than supporting learning. As a result, generated videos often lack explicit pedagogical structure, reliable quality control, and mechanisms for assessing learner understanding or addressing misconceptions. In this work, we introduce PIVOT (Pedagogy-guided Instructional VideO Tutoring), a generative video tutoring framework for STEM learning via learning-centered instructional support. Inspired by conventional teaching workflows, our framework integrates pedagogy into the full generation pipeline: it first uses instructional principles to guide storyboard generation, then produces verified multimodal videos through code-centric generation and a pedagogical verification harness, and finally connects videos with assessment and misconception-aware remediation. Experiments and expert evaluations across four STEM domains show that our framework produces educational videos with pedagogically aligned content, clear and engaging presentation, coherent instructional flow, and perceived effectiveness for learning. These findings suggest a human-centered perspective on educational content generation: generative systems should be evaluated and designed not only by what they produce, but also by how they support teaching practices, learner understanding, and corrective feedback.

As online language education increasingly shifts toward personalized and immersive learning, scenario-based learning (SBL) offers a promising way to connect language knowledge with authentic communicative contexts. However, creating or generating immersive scenario-oriented learning materials remains challenging, since systems must interpret complex real-world situations, organize them into pedagogically coherent learning paths, adapt linguistic difficulty to learners’ proficiency levels, and support interactive learning beyond static content delivery. In this work, we develop ImmersiveCFL, a multi-agent framework that generates scenario-oriented learning paths for enhancing immersive experiences during learning Chinese as Foreign Language (CFL). Given learner-customized scenarios, ImmersiveCFL schedules a pedagogically coherent learning path with knowledge retrieved from a Chinese Education Knowledge Graph, generates multimodal scenario dialogues, and creates task-based interactive simulations with multi-skill exercises. Experimental results show that ImmersiveCFL improves difficulty control, curriculum coherence, and personalization over general LLM-based baselines. An online prototype[http://smartlearn.shuishan.net.cn:8089] further demonstrates its potential for goal-driven and interactive CFL learning.