Hardware Check: Before you install any of the free open-source software reviewed here, stop and check that you have a suitable graphics card.

Some people call this kind of software a "Stable Diffusion interface," but they aren't just for Stable Diffusion anymore. These local interfaces can generate images using a variety of generative AI models. The latest and greatest of generative AI models is called Flux from Black Forest Labs. Reviews start with SwarmUI, Forge, and ComfyUI, three great choices for running Flux locally.

Now being developed at Comfy.org, SwarmUI offers blazing fast speed and great efficiency. It can generate images from a range of models including the latest and most powerful Flux, plus SD1.5, SDXL, and Stable Diffusion 3. Swarm uses ComfyUI as a back-end. You can use Swarm's interface without touching any nodes, but for the technically inclined, it optionally lets you create, edit and run node-based workflows.

The Forge WebUI is growing in popularity. It started as almost a twin of Automatic1111, but with some efficiency changes and different extensions available. After recent updates, it now works with the newest model, Flux, as well as supporting older Stable Diffusion models.

ComfyUI is an advanced node-based interface that can often exceed what’s possible in other packages, while making the most of your computer’s resources. A lot of the newest features tend to come to Comfy first, before they appear in other interfaces. While new users might start with the full SwarmUI suite of tools, Comfy is the technical heart of it, and can still be downloaded by itself, for the people who want to do everything with nodes.

For years, the Automatic1111 WebUI has been the most popular interface to Stable Diffusion for hobbyists and graphics professionals. Packed with options and controls, the Automatic1111 WebUI lets you generate images using a range of available models, use sketches and reference images to guide your generations with ControlNet, modify images with img2img and inpainting, create various kinds of animations, and best of all it lets you use a dazzling array of available extensions.

Easy to install and easy to use, Fooocus starts with a deceptively simple blank screen, and a place to type your prompt. Just type what you want to see, press Generate, and Fooocus starts creating new images right away. But Fooocus isn't just for beginners. It offers many advanced options and features that let you get the most out of available SDXL Stable Diffusion models.

Krita Diffusion is a plug-in to Krita (the powerful open-source paint program) that gives you extensive AI features accessible from inside the paint program. If you want to work directly in a layer-based paint program, but add some AI generated images to selected areas, using entirely free software, then Krita Diffusion could be for you.

InvokeAI is a powerful but carefully designed interface, with bespoke controls for easy regional prompting in areas of your images, a canvas for inpainting and outpainting your images, and other tools. More full-featured than Fooocus but less than Automatic1111, it offers a unique experience that is worth trying for yourself.

SD.Next is an alternative branch of Automatic1111. It starts out with a similar version of the Automatic1111 interface but adds some of its own options and features. A key feature is improved compatibility, so that people with AMD graphics cards should be able to get this running.