Stable Diffusion negative prompt

Categories: Open Source, OpenAI

What is a negative prompt in stable diffusion?

In the context of AI art generation, particularly with models like Stable Diffusion, a "negative prompt" refers to a technique used to guide the AI in what not to include or emphasize in the generated image. It's a part of the prompt engineering process where you explicitly state the elements or attributes you want the model to avoid.

Here's how it works:

Positive Prompts: These are the usual prompts where you describe what you want to see in the image. For example, "a sunny beach with palm trees".

Negative Prompts: These are the instructions added to tell the AI what you don't want in the image. For example, if you add "-crowds, -buildings" to the previous prompt, you're instructing the model to avoid including crowds and buildings in the beach scene.

The combination of positive and negative prompts helps in fine-tuning the output of AI models like Stable Diffusion. This technique is especially useful when you have specific requirements or when you want to avoid common misinterpretations or unwanted elements that the model might otherwise include based on its training data.

Negative prompting is part of the broader field of prompt engineering, which involves skillfully crafting prompts to get the most accurate and desired results from AI models. It's a blend of art and science, requiring an understanding of how the model interprets language and transforms it into visual content.

Negative Prompts for Realism

two person, painting, drawing, anime, CGI, unreal engine, 3d, render, deformed iris

Negative prompts for Anima

(worst quality:1.4), (low quality:1.4), (monochrome:1.1), easynegative,

(worst quality:1.4), (low quality:1.4), (monochrome:1.1), easynegative, bad-artist-anime, bad-image-v2-39000, bad_prompt_version2, bad_quality, ng_deepnegative_v1_75t, verybadimagenegative_v1.1-6400, vile_prompt3,