Claude-real-video － any LLM can watch a video
Summary
Claude-real-video proposes a local, scene-aware video processing pipeline that lets LLMs watch videos by extracting meaningful frames, deduplicating, transcribing audio, and outputting a clean set of frames and transcripts without uploading content. It enables any LLM to interpret video data by providing frames, transcripts, and optional audio locally on the user's machine.