X7xenon/XenClips ? reverse-engineered prompt

Build me a self hosted web app that turns long YouTube videos, podcasts, or local video files into short vertical clips for Shorts, TikTok, and Reels.

I want a clean dark dashboard where I can paste a video link or upload a file, set my API keys in settings, and start a job. The app should download the video, transcribe it, find the most interesting emotional moments, and automatically cut out the best clips. It should also crop the video into a vertical format, add stylish captions with word by word highlighting, and let me preview the results before exporting.

Please make it feel polished and fast, with clear job progress, time estimates, and a place to review and tweak clips after processing. Use a modern React frontend and a Python backend, and keep it fully local so I can run it on my own machine. If you need to check current docs for any libraries, feel free to look them up online.