AIを用いた絵本の自動生成の試み

抄録

We have studied automatic picture book generation using generative neural networks capable of producing both images and text. We selected DALL-E for image generation and GPT-2 for text generation and attempted to generate picture books automatically. However, over the past six months, there has been an explosive proliferation of generative neural networks, exemplified by ChatGPT. In light of this background, we employed GPT-4 and Stable Diffusion in our current study to further explore picture book generation. We confirmed better results in this new experiment.