2021년 MacBook에서 Gemma4-31B로 1년치 영상을 로컬 색인하기(50GB 스왑)

This project demonstrates the feasibility of high-performance local video indexing using Gemma 4-31B on consumer hardware, significantly reducing reliance on expensive cloud SaaS. By leveraging a local-first architecture and sidecar files, it addresses the core bottleneck of media production—searchability—while ensuring data privacy and drastic cost reduction. This shift signals a growing trend where powerful open-weights models enable sophisticated, personalized content management workflows.