ZeyueT/AudioX ? reverse-engineered prompt

Build me a simple app for AudioX that lets me generate audio from text or a video upload, using the pretrained models they mention and a clean Gradio style interface.

I want a page where I can pick a model, paste a prompt, optionally upload a video or reference audio, and then generate and play back the result right in the browser. If the model files are missing, it should download them or clearly tell me what I need. Please make the main run command easy, and keep the experience as close as possible to the README demo.

If you need to, look up any current docs online for the model loading or Gradio bits, but otherwise just get it working end to end with the existing codebase.