Generative Visual Communication in the Era of Vision-Language Models