Humans are critical for the creation and maintenance of high-quality Knowledge Graphs (KGs). However, creating and maintaining large KGs only with humans does not scale, which is especially true for contributions based on multimedia (e.g. images) that are hard to find on the Web and expensive to generate by humans. Therefore, we leverage creative AI for the task of generating images for Wikidata items that do not have them. Our approach uses triples contained in Wikidata and generates natural text descriptions of items about fictional characters with missing images in multiple ways, including verbalizing them through large language models (LLMs). Then, we use those natural text descriptions as prompts for text-to-image models to generate plausible candidate images for Wikidata KG image completion. We design and implement several evaluations to verify the plausibility of our methods.