The easiest free way to make AI images on your own Mac is Draw Things, an app built for Apple Silicon that runs Stable Diffusion and Flux models with nothing uploaded. ComfyUI is the free, open-source choice if you want full control over the workflow. A 16 GB Mac can run Flux at 4-bit; 32 GB is more comfortable. Like local chat models, local image generators keep your prompts and pictures private and work offline once the model is downloaded.
Key facts
- Easiest: Draw Things, free, for Mac, iPhone and iPad.
- Most control: ComfyUI, free and open source (GPL-3.0), for Mac, Windows and Linux.
- Command line: MFLUX runs Flux models on Apple's MLX framework.
- Avoid for new set-ups: DiffusionBee still works, but its public code was last updated in October 2024.
- Memory: Flux at 4-bit needs about 10 GB, so 16 GB works; full-precision Flux wants 64 GB.
On this page
- The best local image apps for a Mac
- What your Mac's memory can run
- How to make your first image
- Can you use the images commercially?
- Frequently asked questions
The best local image apps for a Mac
| App | Price | Best for | Catch |
|---|---|---|---|
| Draw Things | Free | Beginners on Apple devices; fast on Apple Silicon | Mac and iOS only |
| ComfyUI | Free, open source | Full control with node-based workflows, image and video | A learning curve, and slower than Draw Things on a Mac in published comparisons |
| MFLUX | Free, open source | Running Flux from Terminal on MLX | No graphical interface |
| DiffusionBee | Free | Simple one-window app | Public code not updated since October 2024 |
Independent tests published in 2026 found Draw Things about 20 per cent faster than ComfyUI on Apple Silicon. On a Windows PC with an NVIDIA graphics card, ComfyUI is the usual choice.
What your Mac's memory can run
- 16 GB: Stable Diffusion models, and Flux at 4-bit (about 10 GB in use).
- 24 to 32 GB: Flux at 8-bit, with room for larger images.
- 64 GB: Flux at full precision (about 24 GB) with room to spare.
Image generation is slower than chat: expect tens of seconds to a minute or more per large image on a Mac mini, depending on the model and settings. If you are buying a Mac for this, memory matters more than chip speed. See which Mac to buy for local AI.
How to make your first image
- Install Draw Things from the Mac App Store.
- Pick a model in the app. Start with a smaller Stable Diffusion model on 16 GB, or a 4-bit Flux model.
- Type a prompt and generate. The first run downloads the model; after that it works offline.
- Move to ComfyUI later if you want to chain steps, such as upscaling or editing part of an image.
Can you use the images commercially?
It depends on the model's licence, not the app. Some image models allow commercial use and some do not, so read the licence on the model's own page before you use an image for a business. For UK copyright and AI-made images, see who owns AI-generated content?, and for adverts, what the ASA expects.
Frequently asked questions
What is the best local AI image generator for a Mac?
Draw Things for most people: it is free, built for Apple Silicon and runs Stable Diffusion and Flux models. ComfyUI is the open-source choice for full control.
Can a 16 GB Mac generate AI images?
Yes. Stable Diffusion models run well, and Flux runs at 4-bit using about 10 GB. 32 GB gives more room for larger images and better-quality versions.
Can LM Studio or Ollama make images?
Not in the way an image app does. LM Studio and Ollama run text models, some of which read images. To create images locally, use an image app such as Draw Things or ComfyUI.
Is local AI image generation private?
Yes. Your prompts and images stay on your Mac, and the app works offline once the model is downloaded.
Sources
- Independent comparison, Mac Mini local image generation: ComfyUI vs Draw Things
- Independent guide, Local image generation on Mac: what fits your RAM
- Independent guide, Best local AI image generators in 2026 (DiffusionBee repository note)
Apps and models change quickly. Check each project's own page before you install. Spotted something out of date? Tell me.



