HUANGCHIHHUNGLeo/claude-real-video ? reverse-engineered prompt

Build me a simple local tool that lets an AI actually watch a video instead of just reading the transcript.

I want to paste either a video link or pick a local video file, and have it analyze the video on my own machine, not upload the source anywhere. It should pull the frames that really matter when the scene changes, skip near duplicate frames, and also make a speech transcript if audio is available. The result should be easy to hand to Claude, ChatGPT, or Gemini, with exported images, timestamps, a transcript, and a short manifest explaining what was generated.

Please include both a command line version and a small local web page for people who do not want to use the terminal. The web page should let me paste a link or file path, click Analyze, and open a simple viewer showing the video, key frames, and transcript. If possible, support English, Traditional Chinese, and Simplified Chinese.

Keep it cross platform, privacy first, and use current docs online if you need to.