pashpashpash/manga-reader ? reverse-engineered prompt

Build me a Python tool that takes a manga volume PDF and turns it into a short video recap.

I want it to read the pages, pull out the images and important panels, figure out what is happening using GPT 4 Vision, then write a clear story style summary of the volume. After that, use ElevenLabs to turn the summary into narration and combine the narration with the page or panel images into a finished MP4 recap.

It should work from the command line for a manga name and volume number, and save the final video in that volume folder. If it helps, include a simple notebook friendly workflow too so I can run it step by step and debug it more easily. Use the API keys from environment variables, and make the project easy to set up and run with a PDF placed in the expected folder structure.