LPengYang/MotionClone ? reverse-engineered prompt

Build me a working version of this MotionClone project so I can generate videos from a text prompt or an input image while borrowing motion from a reference video.

I want the repo to run cleanly end to end, with simple commands for text to video and image to video, using the sample configs and example files already in the project. Please make sure the setup instructions are clear, the model paths are handled well, and the scripts save generated videos in an easy to find folder. If anything important is missing, like weights or environment details, add helpful checks and friendly error messages so I know what to download or where to put it.

Please keep it focused on the main use cases in the README, especially motion cloning for camera motion, object motion, sketch to video, and image to video. If you need to look up current docs online for any of the external model requirements, go ahead and do that.