TalesRunner

AI-powered storybook generation from user-provided images.

2025.01 - 2025.02

Overview

TalesRunner is a multimodal AI system that transforms user-provided images into a personalized storybook. The system extracts visual descriptions from images, generates story text using a fine-tuned Korean language model, and compiles the final output into a downloadable PDF storybook.

Figure 1. Storybook generator interface. Users can upload images while the system recommends auxiliary story information for personalized storybook generation.

Output

Figure 2. Generated storybook example. Each generated story paragraph is aligned with the corresponding input image and exported as a PDF storybook.

Method

The pipeline combines BLIP-based image captioning, optional story context, KoT5-based story generation, and PDF assembly.

Figure 3. Multimodal storybook generation pipeline. BLIP captions and optional story context are passed to KoT5 and assembled into a PDF storybook.

Tech Stack

Python · BLIP · KoT5 · Transformers · Streamlit · PDF Generation