As the volume of video data grows, efficient retrieval has become crucial for applications in security and multimedia management. Our system, ReViMM, is designed to address this challenge by employing a novel reweighting mechanism that enhances query accuracy. This mechanism dynamically prioritizes essential elements within each query, significantly improving relevance and precision in search results. ReViMM integrates FAISS for similarity search, ElasticSearch for optimized indexing, and Whisper for robust speech transcription, allowing it to process complex multimedia inputs, including text, images, and audio. By combining these tools with large language models to generate precise descriptions, our system delivers a comprehensive solution for accurate, context-sensitive video and multimedia retrieval.