Bilderbuch

Bilderbuch is a hobby project that uses an interactive Streamlit app to generate illustrated books with a focus on local LLMs. The app guides users through topic selection, story outlining, page writing, illustration generation, and PDF export.

Features

• Suggests popular children's book topics using LLMs.
• Generates story outlines and page texts with customizable models.
• Creates hand-drawn style illustrations for each page using Stable Diffusion models.
• Assembles the book into a downloadable PDF.
• All steps are interactive and editable.

Technologies

• Python 3.11+
• Streamlit (UI)
• Ollama (LLM chat API)
• Diffusers, Torch (image generation)
• ReportLab (PDF generation)
• Pydantic (data models)

Installation

1. Clone the repository:

   git clone https://github.com/John-Athan/bilderbuch.git
   cd bilderbuch
   

2. Install dependencies (recommended: use a virtual environment):

   pip install -r requirements.txt
   

   Or use the dependencies listed in pyproject.toml.

3. Make sure you have Ollama running locally for LLM support.

Usage

Run the Streamlit app:

streamlit run app.py

Follow the on-screen steps to create your own illustrated children's book. Download the final PDF when finished.

Project Structure

• app.py: Main Streamlit application.
• model.py: Data models and state management.
• storyline_creator.py: Story and prompt generation.
• image_generator.py: AI image generation.
• pdf_generator.py: PDF creation.
• topic_creator.py: Topic suggestion via LLM.