Diffusers

Diffusers is a cutting-edge library offering pretrained diffusion models for generating videos, images, and audio. Designed for simplicity and flexibility, it allows users to perform easy inference with minimal code and offers the ability to mix-and-match pipeline components like models and schedulers.

The library supports adapters such as LoRA and includes optimizations like offloading and quantization to make large models accessible on devices with limited memory. For those with ample resources, Diffusers enhances performance using torch.compile for faster inference.

Join the Hugging Face community to collaborate on models, datasets, and Spaces. Benefit from accelerated inference and customizable documentation themes. Dive into the world of diffusion models, learn the theory, and start generating or fine-tuning your own models today.