How to use negative prompts in Stable Diffusion

Negative prompts in Stable Diffusion tell the AI what NOT to include in your generated images. They provide precise control over unwanted elements, artifacts, and quality issues that often appear in AI-generated artwork.

1. Locate the negative prompt field. Open your Stable Diffusion interface. Find the negative prompt text box, typically labeled 'Negative prompt' and positioned directly below the main prompt field. In AUTOMATIC1111, click the arrow next to the prompt box to expand the negative prompt section.
2. Start with basic quality negatives. Enter fundamental quality control terms in the negative prompt field: 'blurry, low quality, pixelated, artifacts, distorted'. These universal negatives improve overall image clarity and reduce common AI generation issues across most image types.
3. Add anatomy and body negatives for people. For human subjects, include anatomy-specific negatives: 'extra limbs, extra fingers, malformed hands, missing fingers, duplicate, disfigured face'. Stable Diffusion frequently struggles with human anatomy, making these negatives essential for portrait work.
4. Remove unwanted style elements. Specify style elements to avoid based on your desired output. For realistic images, add 'cartoon, anime, painting, drawing, sketch'. For clean images, include 'text, watermark, signature, logo'. Tailor these to your specific artistic goals.
5. Control unwanted objects and elements. List specific objects or elements you want to exclude from your scene. For outdoor scenes, you might add 'cars, buildings, people' to ensure a natural landscape. For portraits, include 'glasses, hat, jewelry' if you want a clean face without accessories.
6. Use weighted negative prompts for emphasis. Apply weights to negative prompts using parentheses and numbers: '(blurry:1.5)' makes the negative stronger, while '(blurry:0.8)' makes it weaker. Default weight is 1.0. Use weights between 0.5-2.0 for best results without breaking the generation.
7. Test and refine your negative prompts. Generate multiple images with the same positive prompt but different negative prompts to see their effects. Compare results and adjust your negative prompt list based on what consistently improves your outputs. Remove negatives that don't make a visible difference.