AI Based story to Comic genaration using Generative Adversarial Network

Converting a written story into a comic requires the creation of multiple illustrations, consistent characters, suitable scene backgrounds, dialogue placement, and a readable panel sequence. Manual comic creation can be time-consuming and often requires artistic and design skills. This seminar studies an AI-based approach for converting a multi-sentence story into a sequence of visual panels using Generative Adversarial Networks (GANs), with the research paper StoryGAN: A Sequential Conditional GAN for Story Visualization as the primary foundation.StoryGAN treats story visualization as a sequential generation problem. It uses sentence representations, a context encoder, a Text2Gist mechanism, a generator, and image-level and story-level discriminators. These components aim to make generated images relevant to their sentences while considering the wider story context. The proposed project extends the story-visualization idea with preprocessing, sentence/scene segmentation, character and location extraction, prompt or scene-description preparation, and comic-panel arrangement with captions or dialogue.The report explains the core GAN concepts, the StoryGAN architecture, relevant algorithms, related text-to-image research, possible applications, challenges, and future improvements. A key distinction is maintained between the original paper’s image-sequence generation and the additional modules required to produce a formatted comic page. The proposed system is a design for implementation; performance claims will need to be validated through experiments using suitable datasets and evaluation criteria.Keywords: Generative AI, GAN, StoryGAN, story visualization, text-to-image generation, GRU, context encoder, comic generation, NLP.