A design studio uses a foundation model on Amazon Bedrock to generate images for projects. They want to control how detailed versus how abstract each generated image appears. Which model parameter should they adjust?

The generation step parameter (often called "steps" or "inference steps") in image generation models controls the number of iterations the model takes to refine the output image. More steps generally lead to more detailed and higher-quality images, while fewer steps can result in more abstract or less refined outputs, allowing the studio to control the level of detail. Model checkpoint refers to a saved state of the model, not a parameter for real-time adjustment. Batch size determines how many images are generated simultaneously and doesn't directly control detail. Token length is relevant for text generation, not image generation, and refers to the number of output tokens.