AI-Driven Contextual Story Creation with Integrated Text and Visual Generation
Muhammed Gülsoy, Vael Kokach, Büşra Kocaçınar, Fatma Patlar Akbulut · 2024
This study proposes an AI-driven framework for generating coherent narratives and corresponding visuals tailored for children's stories. We produce textual material by using Long Short-Term Memory (LSTM) models, and visually appealing images by using Generative Adversarial Networks (GANs) and fine-tuned Stable Diffusion Models. Furthermore, we investigate various prefix-based story starting points to increase the variety and interest of the story. Our method includes a significant innovation: the Stable Diffusion model is fine-tuned for image generation, enabling the production of high-quality, contextually relevant images from particular story segments. We show by experiments that the Stable Diffusion model can be fine-tuned to greatly improve the quality and relevance of the produced images, and that the layered LSTM model with extended training produces the best text generation performance. Using these advanced models together improved the storytelling experience by fusing imaginative AI with rich, narrative-driven imagery.