hpcaitech/Open-sora ? reverse-engineered prompt

Build me an open source video generation project like Open Sora that can make short videos from text, and also support turning an image into a video, and video editing style workflows if that fits. I want a clean Python codebase with a simple demo interface so I can type a prompt, pick a duration and aspect ratio, and generate a video I can preview and download.

Please include the full pipeline, data preparation, training, and inference so it feels like a real project, not just a demo. Make it practical to run, with clear configs, scripts, and documentation for setup and usage. If there are current best practices or newer docs I should follow, look them up online first.

Keep the project polished and easy to use, with sensible defaults, good error handling, and enough structure that someone can understand how to train or run it without digging through a mess of code.