SDXL Prompt Guide
Structure positive and negative prompts for Stable Diffusion XL, pick CFG/steps confidently, and reuse ready-made templates.
Upload an image → Get SDXL promptPositive + negative prompt templates
Photorealistic portrait
Positive
[subject] portrait, studio photography, soft key light, detailed skin texture, masterpiece, best quality, 8k uhd
Negative
low quality, blurry, bad anatomy, duplicate, watermark, oversharpened
Concept art environment
Positive
[scene], cinematic lighting, volumetric rays, intricate details, trending on artstation, masterpiece, ultra high resolution
Negative
grainy, low contrast, text, watermark, monochrome, distorted
Product render
Positive
premium [product] on reflective surface, studio lighting, clear highlights, photorealistic, 32k render, masterpiece, high detail
Negative
noise, watermark, cluttered background, out of focus, blurry reflections
Recommended settings
Sampler
DPM++ 2M Karras or UniPC
Steps
20-30 for portraits, 35-40 for complex scenes
CFG Scale
7.0-8.0 keeps fidelity without oversaturation
Resolution
1024×1024 square, 1344×768 landscape, 768×1344 portrait
Prompt examples
Editorial fashion shot
Keep negative prompt concise but clear about quality issues.
fashion portrait of a violinist on stage, dramatic spotlights, shallow depth of field, couture styling, masterpiece, best quality, highly detailed, 8k uhd Negative: blurry, low quality, bad anatomy, watermark, duplicate face
Cyberpunk cityscape
Add environment / lighting adjectives before quality boosters.
rain-soaked cyberpunk downtown at midnight, neon reflections, volumetric fog, cinematic composition, masterpiece, ultra detailed, trending on artstation Negative: low contrast, monochrome, text, watermark, noisy shadows
Food photography
Mention lighting type to avoid overly flat results.
gourmet ramen bowl, steam rising, golden hour window light, macro food photography, masterpiece, super detailed, ultra high resolution Negative: oversaturated, watermark, messy background, distorted bowl
SDXL Q&A
How long should the negative prompt be?
Focus on the top 5–7 issues you want to avoid: quality, anatomy, text, duplicates, lighting problems.
Do I need to mention CFG/steps inside the prompt?
No, keep those values in your UI. The prompt should only contain descriptive text.
Why is my image still blurry?
Reduce CFG to 6.5–7, increase steps slightly, and include “sharp focus, detailed texture” in the positive prompt.
Continue Exploring
Generate SDXL-ready prompts automatically
ImgtoPrompt converts any reference image into SDXL positive+negative prompts, then adapts the same description for Midjourney, Flux, and DALL·E too.
Start converting images