About
Mochi Diffusion is a native macOS app for generating images locally on your Mac. It uses Apple's Core ML Stable Diffusion implementation for speed and reduced memory use, and the Iris pipeline for running FLUX.2 Klein models with native Metal GPU acceleration. The app is built with SwiftUI, works completely offline, and requires Apple Silicon (M1 or later) and macOS 15.6 or later.
Generated images are saved as PNG files with their generation settings embedded, which AUTOMATIC1111-compatible tools and Civitai uploads can read. Users can view these settings in Finder's Get Info window. The app includes a built-in gallery that supports importing, saving, and syncing images. Stable Diffusion models can be used for image-to-image generation or with ControlNet, and you can choose compute unit options such as CPU & Neural Engine or CPU & GPU.
Mochi Diffusion requires Core ML Stable Diffusion or FLUX.2 Klein models. Core ML models must be converted or downloaded and placed in a models folder under MochiDiffusion in your home directory, though this location can be changed in Settings. FLUX.2 Klein models need no conversion and can use up to four input images, with the app automatically resizing inputs when needed. The project is licensed under GPL-3.0 and is available in many languages.
What's new
All release notesHomebrew cask mochi-diffusion 6.2 (sha256 b8fcfb33f86fe2d6b34200be717cfba60d8b8f0eb2e2eb4c719bc90420a13042).