用AI自动生成教学视频，减轻教师负担且效果不输人工。

The integration of artificial intelligence (AI) into video lecture production has the potential to transform higher education by streamlining content creation and enhancing accessibility. This paper investigates a semi automated workflow that combines Google Gemini for script generation, Amazon Polly for voice synthesis, and Microsoft PowerPoint for video assembly. Unlike fully automated text to video platforms, this hybrid approach preserves pedagogical intent while ensuring script to slide synchronization, narrative coherence, and customization. Case studies demonstrate the effectiveness of Gemini in generating accurate and context-sensitive scripts for visually rich academic presentations, while Polly provides natural-sounding narration with controllable pacing. A two course pilot study was conducted to evaluate AI generated instructional videos (AIIV) against human instructional videos (HIV). Both qualitative and quantitative results indicate that AIIVs are comparable to HIVs in terms of learning outcomes, with students reporting high levels of clarity, coherence, and usability. However, limitations remain, particularly regarding audio quality and the absence of human-like avatars. The findings suggest that AI assisted video production can reduce instructor workload, improve scalability, and deliver effective learning resources, while future improvements in synthetic voices and avatars may further enhance learner engagement.