<p>This paper introduces a novel approach to personalized image generation by integrating a user-centered recommender system into the image generation pipeline. The proposed method leverages a multimodal recommender system, RecCLIP, which evaluates generated content based on user preferences. This framework establishes a new paradigm at the intersection of recommender systems and generative artificial intelligence. RecCLIP effectively filters generated content in a customized manner to enhance user satisfaction. By utilizing both textual and visual modalities, RecCLIP demonstrates competitive performance in accurately predicting whether a generated image aligns with user expectations. The recommender system can therefore act as a safeguard to filter overwhelming generated content and tailor to user needs. Overall, the implications of this approach lie in producing tailored content for end users and filtering out irrelevant content, thereby delivering targeted results to maximize satisfaction. By combining recommender systems with generative methods, this solution improves alignment with user preferences and enriches the user experience through highly personalized and contextually relevant content.</p>