• Rushika Ravindra Paga
• Siddhesh Bhor
• Pragati Waghmare
• Prajwal Berad

In recent years, artificial intelligence (AI) and multimodal learning technologies have revolutionized the field of e-learning by enabling systems that can understand and generate contextually relevant educational content. However, traditional video-based learning systems still face several limitations, particularly in the areas of information retrieval, personalization, and learner interaction. This research presents an AI-Powered Video Retrieval-Augmented Generation (Video-RAG) system designed specifically for interactive e-learning environments. Inspired by state-of-the-art multimodal retrieval and generative frameworks such as VideoRAG (2025), the proposed model integrates multiple modalities — including video, audio, and textual information — to create a dynamic and adaptive learning experience. The system architecture comprises three core modules: (1) Speech-to-Text Conversion using OpenAI’s Whisper model, which accurately transcribes lecture videos into structured textual data; (2) Multimodal Retrieval utilizing CLIP (Contrastive Language–Image Pretraining) to align visual frames and textual segments for efficient context retrieval; and (3) Generative Response Synthesis, powered by a fine-tuned large language model (LLM), that produces coherent, context-aware answers to user queries. This integration of retrieval and generation capabilities allows students to interactively query lecture content and obtain highly relevant explanations, summaries, or clarifications in real time. The system was evaluated on educational video datasets covering diverse academic domains, including computer science, artificial intelligence, and data science. Quantitative experiments show a significant improvement in retrieval precision, context relevance, and response quality compared to traditional keyword-based or static video learning systems. User studies also reveal enhanced learner engagement, reduced cognitive overload, and higher satisfaction with the system’s adaptive feedback mechanism. Furthermore, the proposed Video-RAG model demonstrates scalability and flexibility, supporting both synchronous and asynchronous learning setups. Overall, this research bridges a crucial gap between multimodal AI systems and interactive e-learning applications by introducing a retrieval-augmented generation framework optimized for long-duration educational videos. The results affirm the potential of Video-RAG architectures to transform digital education by fostering intelligent, context-aware, and personalized learning experiences. The system serves as a foundation for future advancements in AI-driven educational assistants, capable of integrating emotional understanding, adaptive assessment, and domain-specific knowledge generation.

AI, Video-RAG, E-Learning, Multimodal Retrieval, Whisper, CLIP, Interactive Learning

"AI-Powered Video-RAG for Interactive E-Learning", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 11, page no.c321-c330, November-2025, Available :http://www.jetir.org/papers/JETIR2511239.pdf

"AI-Powered Video-RAG for Interactive E-Learning", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 11, page no. ppc321-c330, November-2025, Available at : http://www.jetir.org/papers/JETIR2511239.pdf