OpenX-Inc/flow ? reverse-engineered prompt

Build me a self hosted app that takes a topic and turns it into a finished short video automatically.

I want it to write a narrated script, break it into scenes, generate each scene as AI video clip by clip, then stitch everything together with transitions, subtitles, narration, and music. It should run headless with no UI, support scheduled runs, and be able to do a dry run that only creates the script first. Please make it work with a separate GPU backend service for video generation, and keep it flexible so I can point it at different hosted or self hosted GPU machines. I also want text to speech support with a default voice option and a way to swap in voice cloning later if available.

If possible, include a simple command line flow for generating a video from a topic, plus a scheduler mode for automatic posting. Use current docs online if you need to check anything.