Stable Diffusion Web UI provides a graphical interface for the Stable Diffusion image generation model.
- Stable Diffusion Web UI has no authentication. If someone gets ahold your unique Tor address, they will be able to generate images using your server. They will not be able to see your image generation history.
- This service is in early stage development
- Minimal functionality and some bugs are expected
- This service can use significant system resources. Beware when running it alongside other services that use significant system resources, such as Bitcoin or electrs.
Stable Diffusion is easiest to use in the "txt2img" mode. Enter some text describing your desired image into the "Prompt" box, then press the "Generate" button. Generating an image can take quite a long time, so be patient. Alternatively you can use the "img2img" mode to make changes to an existing photograph or drawing.
There a number of tutorials and guides online to help you navigate the user interface for the first time. For example, stable-diffusion-art.com has a useful overview of the features of both the txt2img and img2img tabs. Effective prompting can also be surprisingly difficult, so if you are unhappy with your results definitely consult a guide on the anatomy of a good prompt. Note that significantly increasing the resolution or batch size may crash the service, requiring you to manually restart it from the Start9 "Installed Services" tab.
All generated images are logged to the stable-diffusion/outputs directory in the File Browser service, and saved images are copied to the neighboring stable-diffusion/saved directory.
The default Stable Diffusion models included with the service (v2.1 and Dreamshaper v8) are generally quite good across a range of different styles and subjects, but you can get better results by choosing a model fine tuned with your specific goals in mind. Civitai is an excellent source of such models. To install one, download your chosen ".safetensors" file and then upload it to the stable-diffusion/models/Stable-diffusion directory in the File Browser service. Note that ".ckpt" files can also work, but are much less secure and may damage your system.
You model should be based on "SD 1.5" or "SD 2.1". No Stable Diffusion XL models are currently supported.
Once the model file is uploaded, you can return to the Stable Diffusion user interface, press the refresh button (🔄) next to the "Stable Diffusion model" dropdown at the top, and then choose your desired model in the dropdown.