The LLM has the video's full context, including visuals. Try chatting with it. Jump to any point in the video and ask questions about what you see.