huggingface/diffusers — reverse-engineered prompt

Build me a Python library for working with diffusion models in PyTorch. I want it to make it simple for someone to load pretrained models, run a ready made pipeline, and generate images from text with just a few lines of code.

Please include a clean modular design so people can use full pipelines when they want something easy, or swap in their own models and noise schedulers when they want more control. It should support the basic ideas behind text to image generation, image generation from noise, and leave room for audio or video pipelines later.

Make it feel like a real open source package, with install setup, clear examples, tests, and friendly documentation. Add a quickstart that shows loading a pretrained model, sending it to GPU when available, running a prompt, and saving the output image. Keep the API simple and readable, and look up current PyTorch and diffusion docs online if you need to.