What we like
- Open models you can run on your own hardware
- Deep control: negative prompts, community-trained styles and add-ons
- Many apps and services build on it, so you can choose your interface
Watch out for
- Setting it up yourself takes technical effort
- Results depend heavily on the model and settings you choose
- Licences differ between model versions, so check before commercial use
Stable Diffusion is a family of image-generation models first released by Stability AI. Many are available openly, which is why a large ecosystem of apps, add-ons and community-trained styles has grown around it.
What it is good at
If you want control, this is the tool. You can run it locally, steer it with negative prompts and extensions, and combine community styles. The trade-off is that you do more of the work.
How prompting works
Describe the subject, setting, light and style, then list what you do not want in the negative prompt. Settings such as steps and guidance change the result, so make small changes and compare.
Who should use it
Stable Diffusion suits tinkerers, developers and anyone who needs a custom workflow. If you want pleasing results with minimal setup, Midjourney or DALL-E may fit better.
Read our guide to negative prompts.