Outline2Narrative: Transforming Story Outlines into Rich Multimodal Narratives

Live Demo

For a live demo, open https://carter-pay-regular-pencil.trycloudflare.com/ on your browser. It is a temporary deployment, so please let me know if the website is down. Recommend to use gemini-2.5 since gemini-3 is easily to get overloaded.

Introduction

Outline2Narrative is an AI-powered story creation system that transforms simple story outlines into rich, multimodal narratives perfect for audiobooks, children's stories, educational content, and creative storytelling projects. The system combines advanced language models with image and audio generation to create complete stories with consistent visuals and engaging narratives.

Unlike traditional AI storytelling systems that generate content incrementally, Outline2Narrative employs a blueprint-first architecture that ensures visual consistency, narrative coherence, and structural completeness from the very beginning. The system enables human-AI co-creation, making story creation easier by providing different story directions at each step, allowing creators to guide, edit, and refine stories at every stage of generation.

Features

• Blueprint-First: Establish canonical world information before generation
• Separation of Concerns: Each agent has a specialized role
• Human-in-the-Loop: Enable editing and refinement at every stage
• Consistency Through Architecture: System design ensures coherence, not just model capability
• Structured Narratives: Pre-planned scene graphs guarantee story completeness
• Story Creation Focus: Options and choices serve to guide story development, making creation easier

Blueprint-First Architecture

• World Bible Creation: Establishes canonical character blueprints, environment templates, and style guides before generating any story content
• Visual Consistency: Maintains character appearance, clothing, and environmental details across all scenes using reusable visual tokens
• Style Preservation: Ensures consistent art style throughout the entire narrative

Multi-Modal Generation

• Text Generation: Creates engaging narrative prose with story direction options to guide narrative development
• Image Generation: Produces scene images that maintain visual consistency using a hybrid reference strategy
• Audio Narration: Generates audio narration perfect for audiobook creation and immersive storytelling

Story Direction Options

• Directional Choices: At each scene, choose from different story directions to explore various narrative paths
• Structured Narratives: Stories follow clear narrative arcs with beginning, middle, and end
• Scene Graph Planning: Pre-planned scene structure ensures story completeness and satisfying conclusions

Human-in-the-Loop Co-Creation

• Text Editing: Edit narrative text and choices at any point
• Image Regeneration: Regenerate images with custom prompts or automatic updates
• Character Customization: Select and regenerate main characters before story begins
• Real-Time Refinement: Iterate on content without restarting the entire generation process

State Management

• Continuity Tracking: Maintains character states, locations, and plot threads across scenes
• World State Updates: Dynamically adds new characters, locations, and plot elements
• Session Persistence: Saves story sessions for later continuation, editing, or export

Installation

Prerequisites

• Python 3.10+ for the backend
• Node.js 16+ and npm for the frontend
• Gemini API Key

Backend Setup

1. Create a conda environment (recommended):

conda create -n outline2narrtive python=3.10
conda activate outline2narrtive

2. Install dependencies:

pip install -r requirements.txt

Frontend Setup

1. Navigate to the frontend directory:

cd frontend

2. Install Node.js dependencies:

npm install

How to Run

Backend Server

1. Start the FastAPI backend server:

cd backend
python main.py

The backend will start on http://localhost:8000 by default.

Frontend Application

1. In a separate terminal, start the frontend development server:

cd frontend
npm run dev

The frontend will typically be available at http://localhost:5173.

Access the Application

1. Open your web browser and navigate to the frontend URL (e.g., http://localhost:5173)
2. On the first page, enter your Gemini API key
3. Once configured, you can start creating your stories!

Quick Deployment to the Internet (Temporary Cloudflare Tunnel)

1. Install cloudflared:

wget https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64.deb
sudo dpkg -i cloudflared-linux-amd64.deb

2. Update frontend/vite.config.ts server block:
   • host: true
   • allowedHosts: ['your-tunnel-host.trycloudflare.com']
3. Restart frontend dev server:

cd frontend
npm run dev -- --host --port 5173

4. Start backend (separate terminal):

cd backend
uvicorn main:app --host 0.0.0.0 --port 8000

5. Start tunnel (third terminal):

cloudflared tunnel --url http://localhost:5173

6. Access https URL printed by cloudflared (your-tunnel-host.trycloudflare.com) while all three processes stay running.

Creating a Story

1. Enter Story Outline: Provide a brief description of your story idea
2. Set Story Goal (optional): Choose from preset goals (Education, Adventure, Horror, etc.) or define a custom goal
3. Configure Story Length: Set the maximum number of scenes (recommand at least 10 scenes to ensure a complete story)
4. Add Reference Images (optional): Upload style reference images to guide visual generation
5. Start Creation: The system will generate the World Bible, create character blueprints, and begin the story

Developing Your Story

• Review Each Scene: Each scene presents narrative text with an accompanying image
• Choose Story Direction: Select from 3 different story directions to explore various narrative paths
• Progress Through Scenes: Navigate through the pre-planned scene graph
• Reach Conclusion: Stories always conclude with satisfying endings

Editing and Refinement

• Edit Text: Click edit buttons to modify narrative text and story direction options
• Regenerate Images: Regenerate scene images with custom prompts or automatic updates
• Regenerate Text: Provide instructions to regenerate narrative content
• Adjust Scenes: Modify the maximum number of scenes during story creation
• Export Story: Save your completed story as a JSON file for sharing, audiobook production, or later use

Load Your Created Story

After exporting your story as a JSON file, you can reload it later by using the "Load Story" button on the main page to review.

Use Cases

• Audiobook Creation: Generate complete narratives with audio narration ready for audiobook production
• Children's Stories: Create engaging, illustrated stories with consistent characters and settings
• Educational Content: Develop structured educational narratives with specific learning goals
• Creative Writing: Use story direction options to explore different narrative paths and refine your story
• Content Production: Generate multimodal content (text, images, audio) for various storytelling projects
• Video Story Production: Use consistent story narratives and images as the foundation for creating video stories and animated narratives