Mochi Diffusion

Mochi Diffusion is a macOS native application that runs Stable Diffusion and FLUX.2 Klein models directly on Apple Silicon hardware. By leveraging Apple’s Core ML implementation, it executes inference on the Neural Engine or GPU, achieving fast generation with low memory usage (around 150 MB). The app supports image‑to‑image, ControlNet, and custom Core ML models, and it stores prompt information in EXIF metadata of generated files.

The software is aimed at users who want to generate images locally without internet access, including artists, designers, and developers working on Mac computers. Its SwiftUI interface provides a built‑in gallery with import, save, and sync capabilities, while the model management system lets users add or convert models in a simple folder hierarchy.

Distinctive aspects include full offline operation, efficient use of the Neural Engine for speed and memory savings, and support for both Stable Diffusion and FLUX.2 Klein pipelines. The app also embeds prompt data in image metadata and offers flexible compute‑unit choices (CPU + Neural Engine or CPU + GPU) to match different Mac configurations.