An easy-to-follow guide for getting started with A1111’s Stable Diffusion WebUI.
⚠ NOTE :
This guide is designed for users who have attended the AI for Artists - Basics Workshop.
Text to Image
Prompts
In the Prompt textbox, describe the image that you would like to generate.
In the Negative Prompt to describe the elements that you would like to avoid in your image.
Styles
Download the styles.csv file and copy the file to :/stable-diffusion-webui/
Select the style of your choice to significantly enhance the quality of your generated images. Combine multiple styles together for interesting results.
Working with Custom Models / Checkpoints
Using custom models from the internet can greatly enhance the quality of your images.
Download a model of your choice and copy the file to :/stable-diffusion-webui/models/Stable Diffusion/
Recommended models :
⚠ NOTE :
Please ensure safety of your system by downloading models from trusted sources that are reviewed by other users.
Resolution
Select the resolution of your image. For best results, refer the following table :
| Aspect Ratio | Resolution |
|---|---|
| 1:1 | 512 x 512 |
| 4:3 | 683 x 512 |
| 3:2 | 768 x 512 |
| 16:9 | 910 x 512 |
| 2:1 | 1024 x 512 |
Seed
To regenerate the same image, copy the seed value from the previously generated image of your choice. You can find the seed value of any image in the filename, or by loading it in the PNG Info tab. When Seed is set to -1, a random image will be generated every time.
CFG Scale
Classifier Free Guidance (CFG Scale) is used to control the creativity of the AI. Higher value will force the AI to strictly follow the prompt as you have described above, while lower value allows more flexibility to be creative.
Recommended range : 3 - 10
Sampling Steps
Sampling steps directly affect the speed and quality of your image generation. Start with lower values for quick explorations. Progressively increase the value to get better results.
Recommended range : 10 - 30
Sampling Methods
Choosing from a long list of available sampling methods can be difficult. Experiment with different options to get variations for the same prompt.
Image to Image
Image Input
Use an image of your choice to guide the generation of your image.
⚠ NOTE :
Scale down your input images to the recommended resolutions before use.
Denoising Strength
This is the most important slider in any Image to Image workflow. Use this slider to control the amount of noise to be added to the input image.
Use Lower values to add details to an existing image without losing the likeness.
Use Higher values to generate entirely new images based on the input image.
Resize
Enlarge your image to a higher resolution while adding details to your generated image.
Inpaint
Regenerate specific parts of the images with higher level of detail.
Inpaint Sketch
Use the color selector to specify the color of your regenrated image.
Extras
Upscale
Enlarge your generated images using an AI Upscaler model. Recommended models :
- ESRGAN
- R-ESRGAN
- R-ESRGAN Anime+
Installing Upscaling Models
External Upscaling models can be found on the internet.
Download a model of your choice and copy the file to :/stable-diffusion-webui/models/ESRGAN/
Recommended models :
⚠ NOTE :
Please ensure safety of your system by downloading models from trusted sources that are reviewed by other users.
PNG Info
Drag any generated image to the PNG Info tab to access the parameters used to generate that image.
Extensions
Aspect Ratio Selector Plus
An extremely helpful extension to quickly switch between different aspect ratios.
Updates
Updating the WebUI
Update your WebUI to the latest version by following the instructions in the video :