Instead of Midjourney

Image generation from a prompt.

Partly

Running Stable Diffusion or Flux locally is real and the results are good. What you lose is the taste: Midjourney's defaults are better than any raw model's.

You lose: Aesthetic defaults, and a GPU you do not have to think about.

Start here

ComfyUI

GPL-3.0a weekendactive

A node graph, and a vocabulary to learn. Extremely capable once you do.

github.com/Comfy-Org/ComfyUI www.comfy.org

A database, object storage or mail config, plus backups.

Also worth a look

InvokeAIApache-2.0an eveningactive
A friendlier canvas over the same models. Fewer exotic techniques.
github.com/invoke-ai/InvokeAI invoke-ai.github.io/InvokeAI
Stable Diffusion WebUIAGPL-3.0an eveningactive
The one everyone's tutorial assumes. Development has slowed.
github.com/AUTOMATIC1111/stable-diffusion-webui

Facts checked 8 November 2026. Licences, activity and archive status come from each project's own repository; the verdict and the cost lines are the desk's.