HVision-NKU/StoryDiffusion ? reverse-engineered prompt

Build me a small app for generating a story comic from a series of text prompts, where the main character stays consistent across every panel.

I want to paste in at least 3 prompts, ideally 5 or 6, and get back a clean set of matching images that feel like one story. It should also have a simple local demo interface so I can run it on my own machine and try different prompts easily. If it makes sense, add a low memory mode so it can run on a smaller GPU without crashing. Please include a way to save the generated images, and make the interface easy enough that I can test it from a notebook or a browser. If you need to check current docs online for anything, go ahead.