Mochi Diffusion
A local AI image generator runs Stable Diffusion and FLUX.2 Klein fully offline on Apple Silicon Macs, with ControlNet and image input support.
Alternative to:
Mochi Diffusion is a native macOS app for people who want to generate images on their own Mac. Generation works fully offline, and the app sends no generation data to the cloud or collects telemetry.
- Stable Diffusion supports community models converted to Core ML, with the option to generate from a starting image or use ControlNet to guide the result.
- FLUX.2 Klein runs the distilled 4B and 9B models without conversion and accepts multiple input images, automatically resizing them when necessary and showing the adjusted dimensions.
- Image gallery lets you import, save and sync images within the app.
- Saved generation settings stay attached to generated PNGs and can be read by AUTOMATIC1111-compatible tools and Civitai uploads, or viewed in Finder's Get Info window.
The app requires macOS 15.6 or later and an Apple Silicon Mac (M1 or later). Stable Diffusion can use the Neural Engine for lower memory use, or CPU and GPU processing, which may be faster on some Macs but uses more memory. Each converted Stable Diffusion model has a fixed output size. FLUX.2 Klein uses Metal GPU acceleration.
Mochi Diffusion is open source under GPL-3.0.
Logotyper
