Negative prompting is a prompt engineering technique that specifies elements, characteristics, or attributes that should be excluded from generated content.
In image generation, negative prompts help guide the model away from producing unwanted objects, visual features, or stylistic characteristics.
For example, a company generating promotional images could specify negative prompt attributes such as watermarks, text overlays, distorted objects, or unwanted backgrounds.
Amazon Bedrock supports image generation models that provide negative prompting capabilities. Certain Stability AI image models expose a negative_prompt parameter for specifying undesirable visual elements.
Zero-shot prompting involves asking a model to perform a task without providing examples. Experimentation prompting and discovery prompting do not specifically describe excluding image attributes.
Negative prompting is therefore the appropriate technique, although generated images should still undergo validation because negative instructions cannot guarantee complete exclusion.