Diffusers

Diffusers is Hugging Face's library for diffusion models, built around three interchangeable parts: pretrained pipelines (run Stable Diffusion-class image generation, audio, or 3D molecule models in a few lines), schedulers (swap noise schedules to trade speed against quality), and models (UNet/transformer building blocks for custom systems). Its stated philosophy is usability over raw performance and tweakability over abstraction, which made it the default way researchers and builders experiment with generative media models. Apache-2.0, with 4.97M PyPI downloads/month.

```bash

pip install diffusers accelerate

```

```python

from diffusers import DiffusionPipeline

import torch

pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=torch.float16)

pipe.to("cuda")

image = pipe("a photo of a cat wearing a hat").images[0]

image.save("cat.png")

```

Docs: huggingface.co/docs/diffusers. (Model availability depends on each checkpoint's own license.)