Setting Up SDXL with Automatic1111

Prerequisites for Using SDXL

To get started with SDXL in Automatic1111, you'll need:

• A working installation of AUTOMATIC1111 (compatible with Windows 10/11 and Mac M1/M2 chips)
• A powerful computer, ideally with an 8GB GPU or better
• If your hardware is lacking, consider cloud-based solutions

Updating Your Existing Installation

If you installed the WebUI before July 24, 2023 (pre-version 1.5.0), update it:

1. Navigate to your installation directory (\stable-diffusion-webui)
2. Run the command: git pull
3. The update should complete quickly

Adding SDXL 1.0 to Automatic1111

Follow these steps for both Mac and Windows:

1. Download the base model and Refiner
2. Place both files in /stable-diffusion-webui/models/Stable-diffusion
3. Launch WebUI or refresh the model list
4. Select the new models (sd_xl_base and sd_xl_refiner)

Download Links:

• SDXL 1.0 Base Model on HuggingFace
• SDXL 1.0 Refiner on HuggingFace

Using SDXL

Basic Text-to-Image Generation:

1. Choose SDXL 1.0 base model from the Stable Diffusion Checkpoint menu
2. Input your prompt (and optional negative prompt)
3. Set image size to 1024x1024 or similar ratios
4. Click Generate

Example Prompt:
"A hyper-realistic lion in a suits looking serious. peaky blinders environment."

Negative Prompt:
"sketch, ugly, blur, text, logo, monochrome, bad art, cut off, low contrast, underexposed, overexposed, beginner, amateur"

Settings: 1024x1024, CFG Scale: 7, Steps: 20, Sampling Method: Euler A

Applying the Refiner in Text-to-Image

With Automatic1111 version 1.6.0 or later:

1. Click "Refiner" under Sampling Method
2. Select the Refiner checkpoint
3. Set when the Refiner activates (e.g., 0.8 for 16 base steps, 4 Refiner steps in a 20-step generation)

Using the Refiner in Image-to-Image

To enhance a generated image:

1. Send your image to img2img
2. Select the Refiner model
3. Adjust denoising strength (0.1 to 0.3)
4. Keep prompts unchanged and generate

Experiment with different denoising strengths for varied results.

Utilizing Predefined Styles

Install StyleSelectorXL:

1. Open the Extensions tab
2. Choose "Available"
3. Load from the default URL
4. Find and install "Style Selector for SDXL"
5. Apply changes and restart the UI

Using StyleSelectorXL:

1. Find the SDXL Styles window in img2img settings
2. Enable Style Selector
3. Choose a style
4. Generate your image

Experiment with various styles like Anime, Stained Glass, Retro Arcade, Disco, Grunge, and Film Noir.

Advancing with SDXL and Automatic1111

Recent updates make SDXL as user-friendly as earlier versions, with improved rendering and high-resolution capabilities. While technically advanced, SDXL's prompt understanding remains similar to previous models.

For inspiration, explore prompt engineering tutorials or use AI assistants to help craft portrait prompts. Before diving in, check out sample prompts for SDXL and follow relevant social media accounts for tips and new developments.