Analyze a local or downloaded video through timestamped human/automatic captions, ASR fallback, uniform and targeted ffmpeg frames, contact sheets, sanitized metadata, and a claim-evidence ledger. Use when Codex must summarize, compare, fact-check, extract rankings or recommendations, interpret slides/gameplay/demos, or create a Markdown/HTML report from Bilibili, YouTube, or local video while keeping conclusions traceable to exact speech and visual evidence.