Background
  1. My favorite BiliBili/Xiaohongshu/Douyin videos are difficult to quickly review
  2. The collection is just a link to the collection. If the author removes the video, the collection will naturally be lost
Requirements
## Workflow
  1. Favorite videos (BiliBili, Xiaohongshu, Douyin, YouTube, etc.)
  2. Automatically extract word-by-word transcripts of conversations (including speakers and speaking time) (see Feishu Miao Ji)
  3. Automatically capture key images (board books/notes/key operation steps) in the video and display them according to the timeline and word-by-word script
    mixed graphic
    or
    side by side
  4. AI summarizes the video content and keeps the summary in the card, making it easy to find
## Another way to implement
  1. Supports markdown mixed graphics notes
  2. The collection API supports the delivery of markdown mixed graphics notes
  3. After the user manually collects the analysis video, the analysis results are directly passed to CuBox using the API
·