Faster Stable Diffusion with Core ML on iPhone/iPad/Mac

AI Impact Summary

Apple has accelerated Stable Diffusion inference on iPhone, iPad, and Mac by leveraging Core ML’s on-device machine learning capabilities. This is achieved through 6-bit palettization, which dramatically reduces model size and memory usage, coupled with optimizations for the Neural Engine. Developers can now run Stable Diffusion models directly on Apple devices, offering benefits like convenience, privacy, and reduced API costs, particularly with the updated ml-stable-diffusion repo and available Core ML models.