Diffusers 0.3 adds Stable Diffusion Img2Img, Inpaint, Textual Inversion, and ONNX export

AI Impact Summary

Diffusers 0.3 expands diffusion-model tooling with new pipelines for Stable Diffusion, including image-to-image generation via StableDiffusionImg2ImgPipeline and inpainting via StableDiffusionInpaintPipeline, plus Textual Inversion for personalizing concepts. It also introduces an experimental ONNX pathway with StableDiffusionOnnxPipeline and a checkpoint-to-ONNX exporter, enabling deployment on non-PyTorch runtimes and CPU/GPU hybrids. The release also highlights memory optimizations and MacOS MPS support, lowering hardware barriers and broadening the potential deployment surface for diffusion-based applications.