Different robotics tasks require different robot morphologies and visual designs. The growing number of newly developed robots demands the creation of innovative and diverse visual designs, especially for those used in socially interactive tasks. Creating innovative robot designs is challenging and relies on highly skilled designers in the classical design process, which can be expensive and time-consuming. Design requirements must align not only with functionality but also with the domain and end-users. However, the influence of end-users and domain experts on the final design is limited, as they primarily communicate requirements rather than directly contribute to the design process. A new design process is proposed where generative artificial intelligence is used to create the visual design of robots. The new process enables iterative refinement of designs and exploration of novel ideas with generative artificial intelligence. A tool realizing the new design process with generative artificial intelligence is developed that enables the creation of novel robot appearances while relying less on design skills. End-users can bring design ideas to life, while designers also benefit from rapidly generating and exploring new concepts. The process supports exploration with minimal guidance and provides detailed user control to refine designs. To improve domain-specific guidance Low-Rank Adaptation is used to fine-tune Stable Diffusion for robot appearances. By using generative artificial intelligence in an iterative process, robot designs can be systematically improved and changes can be tracked. Additionally, approaches for merging resulting designs through generative artificial intelligence are explored, which enables more users to influence the resulting robot’s appearance.