How to Use SDXL: Beginner's Tutorial

Learning how to use SDXL opens access to the most customizable AI image generator available. Stable Diffusion XL combines excellent base quality with an unmatched ecosystem of fine-tuned models, LoRAs, and tools. This beginner’s tutorial covers everything from first generation to advanced techniques.

What is SDXL?

Stable Diffusion XL is Stability AI’s flagship open-source image generation model. Released in 2023, it features:

• 1024x1024 native resolution
• Improved prompt understanding
• Better composition handling
• Two-stage generation process (base + refiner)
• Massive community ecosystem

Quick Start Options

Method	Difficulty	Cost	Best For
Fooocus	Easiest	Free	Beginners
Web Services	Easy	Free-Paid	Quick access
Automatic1111	Medium	Free	Control
ComfyUI	Advanced	Free	Maximum flexibility

Platform Comparison

Feature	SDXL Alone	Multic
AI Images	Yes	Yes
AI Video	Limited	Yes
Comics/Webtoons	No	Yes
Visual Novels	No	Yes
Branching Stories	No	Yes
Real-time Collab	No	Yes
Publishing	No	Yes

Method 1: Fooocus (Recommended for Beginners)

Fooocus provides a Midjourney-like experience with SDXL.

Installation

1. Download Fooocus:

   • Windows: Download from GitHub releases
   • Mac/Linux: Clone repository

2. Run:

   • Windows: Double-click run.bat
   • Mac/Linux: Run launch script

3. First Launch:

   • Fooocus downloads SDXL automatically
   • Wait for model download (several GB)
   • Browser opens to interface

Using Fooocus

1. Enter your prompt in the text box
2. Click Generate
3. Wait for results (30 seconds to 2 minutes)
4. Download or continue iterating

That’s it. Fooocus handles settings automatically.

Fooocus Tips

• Use the “Quality” preset for best results
• Try “Speed” preset for faster iteration
• Advanced tab reveals more control when ready
• Styles tab adds artistic directions

Method 2: Web Services

Clipdrop

1. Visit clipdrop.co/stable-diffusion
2. Enter prompt
3. Generate (free tier available)

Playground AI

1. Visit playground.ai
2. Select SDXL model
3. Generate (500 free daily)

Leonardo AI

1. Visit leonardo.ai
2. Create account
3. Select SDXL-based model
4. Generate (150 free tokens daily)

Tensor.Art

1. Visit tensor.art
2. Choose SDXL model or checkpoint
3. Generate (100 credits daily)

Method 3: Automatic1111 WebUI

The most popular SDXL interface with extensive features.

System Requirements

• GPU: 8GB+ VRAM (NVIDIA recommended)
• RAM: 16GB+
• Storage: 30GB+ for models
• Python 3.10

Installation

Windows (Easiest):

1. Download webui.zip from GitHub releases
2. Extract to desired location
3. Run webui-user.bat
4. Wait for automatic setup

Manual Installation:

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui
cd stable-diffusion-webui
./webui.sh  # Linux/Mac
webui-user.bat  # Windows

First Run Setup

1. First launch takes time (downloads dependencies)
2. Browser opens to localhost:7860
3. Interface loads with default model
4. Download SDXL model separately if not included

Downloading SDXL

1. Get sd_xl_base_1.0.safetensors from Hugging Face or Civitai
2. Place in stable-diffusion-webui/models/Stable-diffusion/
3. Refresh or restart
4. Select from model dropdown

Basic Generation

1. Select SDXL model from dropdown
2. Enter prompt in first text box
3. Enter negative prompt (optional)
4. Set dimensions to 1024x1024
5. Keep sampling steps at 20-30
6. Click Generate

SDXL Prompting Guide

Basic Structure

[subject], [description], [style], [quality tags]

Example:

a majestic lion, golden mane flowing in wind, standing on rocky outcrop at sunset, digital painting, highly detailed, dramatic lighting

Positive Prompt Tips

Be Specific: “woman with red hair” vs “a young woman in her 20s with vibrant red curly hair, pale skin, green eyes”

Include Style: Always specify art style—“oil painting,” “anime style,” “photorealistic,” “watercolor”

Add Quality Tags: “highly detailed,” “professional,” “masterpiece,” “sharp focus”

Describe Lighting: “dramatic lighting,” “soft natural light,” “golden hour,” “studio lighting”

Negative Prompts

SDXL benefits from negative prompts. Common negatives:

ugly, blurry, low quality, worst quality, deformed, disfigured, extra limbs, bad anatomy, watermark, signature, text, cropped

Prompt Weighting

Emphasize or de-emphasize terms:

• (important:1.3) - More emphasis
• (less important:0.7) - Less emphasis
• ((very important)) - Double brackets = 1.1 weight each

Working with SDXL Checkpoints

The SDXL ecosystem includes thousands of specialized models.

Popular Checkpoints

Photorealistic:

• RealVisXL - Best for realistic humans
• Juggernaut XL - Versatile realism

Artistic:

• DreamShaper XL - Fantasy and creativity
• SDXL Unstable Diffusers - Artistic styles

Anime:

• Animagine XL - High quality anime
• Pony Diffusion - Versatile anime/furry

Installing Checkpoints

1. Download .safetensors file from Civitai or Hugging Face
2. Place in models/Stable-diffusion folder
3. Refresh model list in interface
4. Select new checkpoint

Using LoRAs

LoRAs add specific styles, characters, or concepts.

Finding LoRAs

• Civitai.com - Largest collection
• Hugging Face - Official and community uploads

Installing LoRAs

1. Download .safetensors LoRA file
2. Place in models/Lora folder
3. Refresh interface

Using LoRAs

In prompt, add:

<lora:lora_name:strength>

Example:

a fantasy castle, <lora:fantasy_architecture:0.8>, detailed, epic

Strength typically 0.5-1.0.

ControlNet Basics

ControlNet provides precise composition control.

Common ControlNet Types

• Canny Edge: Follow line art or edges
• Depth: Match depth map for composition
• OpenPose: Control human poses
• Tile: Upscale or add detail

Using ControlNet

1. Install ControlNet extension
2. Download ControlNet models
3. Upload control image
4. Select control type
5. Generate with controlled composition

Why Multic is Better for Creators

SDXL generates individual images. Story creators need complete workflows.

Integrated Storytelling: Multic combines AI image generation with comics, visual novels, and interactive narratives.

Character Consistency: Maintain consistent characters throughout your story without managing LoRAs manually.

AI Video: Multic includes video generation—SDXL alone doesn’t offer this.

Collaboration: Work with other creators in real-time.

Publishing: Share completed stories directly, not just exported images.

Creative Need	SDXL Alone	Multic
Generate Images	Yes	Yes
Generate Video	Limited	Yes
Character Consistency	Manual LoRAs	Automatic
Build Comics	Manual	Integrated
Visual Novels	Manual	Integrated
Interactive Stories	No	Yes
Collaborate	No	Yes
Publish	No	Yes

Common SDXL Issues

Out of Memory

• Lower resolution to 768x768
• Enable lowvram or medvram mode
• Reduce batch size to 1

Slow Generation

• Use smaller resolution for drafts
• Reduce steps (20 often enough)
• Try SDXL Turbo for speed

Bad Anatomy

• Add negative prompt for deformities
• Use better checkpoint
• Try ControlNet for poses

Blurry Results

• Increase steps (25-30)
• Use “sharp focus” in prompt
• Try different sampler

Recommended Settings

Standard Quality

• Resolution: 1024x1024
• Steps: 25
• Sampler: DPM++ 2M Karras
• CFG Scale: 7

Fast Drafts

• Resolution: 768x768
• Steps: 15
• Sampler: Euler A
• CFG Scale: 7

Maximum Quality

• Resolution: 1024x1024
• Steps: 40
• Sampler: DPM++ 2M SDE Karras
• CFG Scale: 7
• Use refiner model

The Verdict

SDXL offers unmatched customization and zero ongoing cost. The ecosystem of checkpoints, LoRAs, and tools enables any artistic style imaginable.

For individual image generation, mastering SDXL provides capabilities that rival or exceed paid services. For storytellers building complete narratives, consider whether assembling separate images into stories manually serves your goals—or whether integrated platforms like Multic provide a more effective creative workflow.

---

Ready to use AI images for storytelling? Start on Multic with integrated creative tools.

---

Related: SDXL vs Midjourney and SDXL Turbo vs SDXL