CLIPMIND: The Intelligent Video Shrinker

Abstract

The rapid expansion of long-form digital video content has created a growing need for efficient methods to extract essential information without requiring full-length viewing. ClipMind is a speech-guided video summarization framework that transforms lengthy recordings into compact, semantically coherent highlight videos. Rather than depending solely on visual keyframe detection, the framework derives importance directly from spoken content through automatic speech recognition, topic-aware sentence ranking, and transformer-based abstractive summarization. The generated summary is aligned with time-stamped transcript segments to enable precise extraction and reconstruction of context-preserving video highlights. By jointly integrating transcript analysis and segment mapping within a unified pipeline, the system maintains narrative continuity while significantly reducing video duration. Evaluation on educational and lecture-style content demonstrates effective compression with strong semantic retention and improved user comprehension. The proposed framework is well suited for applications in e-learning, professional training, content indexing, and rapid media consumption scenarios where efficient knowledge extraction is essential.