The Best Ways to Run Stable Diffusion and SDXL on an Apple Silicon Mac
The go-to image generator for AI art enthusiasts can be installed on Apple's latest hardware.
Jose Antonio Lanz8 min read
Mochi Diffusion is a native implementation for running Stable Diffusion models compatible with CoreML. It does not offer an option to convert models, but there is a catalog powered by the Hugging Face community for those who want to use this implementation. The split-einsum models are the best for Apple Silicon, although it also supports general models for those who want some flexibility in choosing to render with a GPU, CPU, and achieving good results at different resolutions.
It is installed via the project's official repo, and the customized models for Apple Silicon can be downloaded from Hugging Face and copied into the corresponding folder.
Conclusion
There is no such thing as a perfect app, and AI artists tend to run different implementations to harness the best of each one. Considering the struggles of having a Mac in a PC-dominated world, you might consider taking the leap and have a few apps installed —just like most of the AI artists out there.
Choosing the right Stable Diffusion implementation for your Apple Silicon Mac depends on what kind of user you are and what you want to accomplish. But with the options presented above, you can surely find something that fits your needs.