Curated, ready-to-use Stable Diffusion prompts across 8 categories. Copy any prompt, paste into your SD interface, or use our AI to improve it further.
✨ AI Prompt Improver
Paste any SD prompt → get a more detailed, optimized version instantly
Pro tip: Use our AI Prompt Improver to add quality boosters, lighting details, and camera specs to any of these prompts automatically. Just click "Improve with AI" on any card below.
📷 Photorealistic Prompts
Golden Hour Portrait
photo of a beautiful woman with flowing auburn hair, golden hour sunlight, bokeh background, shallow depth of field, Canon 85mm f/1.4, soft skin texture, natural makeup, outdoor garden setting, award-winning photography, 8k resolution, highly detailed
Street Food Market
busy night market in Bangkok, steam rising from street food stalls, neon signs in Thai script, candid photography, dramatic lighting, documentary style, shot on Fujifilm X-T4, warm color grading, crowd of locals and tourists, 35mm lens, photorealistic, highly detailed
Mountain Lake Reflection
crystal clear alpine lake reflecting snow-capped mountains at dawn, perfectly still water mirror, early morning mist, wildflowers on the bank, long exposure photography, National Geographic quality, ultra-sharp, HDR, 16:9 composition, photorealistic, 8k
Urban Architecture
modern glass skyscraper facade, geometric patterns of windows reflecting blue sky and clouds, architectural photography, shot looking upward, symmetrical composition, leading lines, Nikkor 24mm tilt-shift lens, sharp focus, commercial real estate photography quality
Wildlife Close-Up
close-up portrait of a Bengal tiger in tall grass, eye contact with camera, shallow depth of field, catchlight in eyes, perfect fur texture detail, golden afternoon light, wildlife photography, 600mm telephoto, National Geographic cover quality, 8k
Rainy City Night
rain-soaked city street at night, neon reflections on wet asphalt, lone figure with umbrella, cinematic color grading, film photography aesthetic, Kodak Portra 400 emulation, moody urban atmosphere, 50mm lens, bokeh streetlights, photorealistic
Underwater Diver
scuba diver exploring a coral reef, rays of sunlight penetrating crystal blue water, colorful tropical fish surrounding the diver, underwater photography, wide angle lens, perfect water clarity, Great Barrier Reef, photorealistic, highly detailed, 8k
Macro Flower
extreme macro photograph of a rose in full bloom, water droplets on delicate petals, razor-sharp focus at center, creamy bokeh background, studio lighting, fine art photography, vibrant red against soft background, 100mm macro lens, 8k resolution
Essential Negative Prompts
Copy these into your negative prompt field to consistently improve output quality:
photo, photograph, realistic, photorealistic, real person, camera, lens flare, HDR
Recommended Stable Diffusion Settings
Sampler
DPM++ 2M Karras
Best balance of speed and quality
Steps
20–30
20 for speed, 30 for detail
CFG Scale
7–9
Lower = more creative, Higher = more literal
SDXL Resolution
1024×1024
Or 832×1216 for portrait
SD 1.5 Resolution
512×512
Use hires.fix to upscale
Hires Denoising
0.3–0.5
Lower preserves composition
Frequently Asked Questions
What makes a good Stable Diffusion prompt?
A strong SD prompt has several layers: subject (what you want), style (photorealistic, oil painting, anime, etc.), quality boosters (highly detailed, 8k, masterpiece, trending on artstation), lighting, composition, and optionally a camera/lens specification for photorealistic work. Keep it dense with descriptive keywords — SD responds to keyword density differently from natural language.
What are the best quality booster words for Stable Diffusion?
Popular quality boosters include: "highly detailed", "8k resolution", "masterpiece", "best quality", "trending on artstation", "sharp focus", "award-winning photography", "photorealistic", "ultra-detailed", "cinematic lighting". Use negative prompts like "blurry, low quality, bad anatomy" to exclude common issues.
What is a negative prompt in Stable Diffusion?
A negative prompt tells Stable Diffusion what to avoid. Common negatives: "blurry, low quality, bad anatomy, extra limbs, watermark, signature, poorly drawn, deformed, ugly, grainy, pixelated, overexposed". For portraits specifically add: "bad hands, extra fingers, missing fingers, distorted face".
Which Stable Diffusion model should I use?
It depends on your use case. SDXL 1.0 is the flagship for photorealism and general use. Juggernaut XL and RealVisXL excel at photorealistic portraits. DreamShaper XL and Proteus are great for artistic work. For anime specifically, AniVerse or Animagine XL are popular choices. Specialized checkpoints on CivitAI often outperform base models for specific styles.
What are the best Stable Diffusion samplers?
DPM++ 2M Karras is the most popular for its balance of speed and quality. Euler A gives varied results good for creativity. DPM++ SDE Karras produces sharp, consistent results. For most use cases: DPM++ 2M Karras at 20-30 steps is a great starting point.
What resolution should I use in Stable Diffusion?
For SD 1.5 models, 512x512 or 512x768 is the native resolution — going higher requires hires.fix. For SDXL models, 1024x1024 is native. For portraits use 832x1216 (portrait), landscapes 1216x832 (wide), and square subjects 1024x1024. Always enable hires.fix if you want final images above the native training resolution.
How do I get consistent characters in Stable Diffusion?
Use ControlNet with a reference image and the "Reference" or "IP-Adapter" preprocessor to maintain character consistency. For a character you own, train a LoRA or use Dreambooth fine-tuning (10-20 images is enough). In the prompt, use distinctive physical features (hair color, eye color, distinctive marks) in every generation.
Can I use Stable Diffusion images commercially?
It depends on the model license. SDXL is open-source under CreativeML Open RAIL+M, which allows commercial use with some restrictions. Check each specific model's license on their respective download page or HuggingFace — some community checkpoints have non-commercial restrictions. Always review the license for any model you use commercially.