GenAI for Little Readers : Personalized Storybook Creation with Generative AI

Hello everyone,

It's been a long time since I was last active here, hasn't it?
How have you all been?

5-Day Gen AI intensive Course with Google x Kaggle

From March 31 through April 4, 2025 I  joined a 5-Day Gen AI intensive Course with Google and Kaggle. Thanks to my friend, I can join such a great free course where Google Experts go through foundational gen AI topics. Even though the time difference between Korea (Korean Standard Time, KST) and Pacific Time (PT) is 16 hours, I still can joined the recorded daily livestreams. 

I've decided to join the Gen AI Intensive 2025 Capstone project, where I have to use what we learned in the project. You can view my project here GenAI Capstone 2025q1 project

Gen AI intensive Capstone Project : Personalized Storybook Creation with GenAI

Title : GenAI for Little Readers

While doing this project, I explored how Generative AI can be used to create personalized storybooks for children — combining creative storytelling and visual imagination. My inspiration for this project comes from my mom, who runs a kindergarten called Little Abqari Playschool. I always helped my mom in the kindergarten whenever I go back home for semester break.

During preschool convocation last year

Help with the parents registration

This was when I joined my mom preschool trip      

We always brainstorming ideas together to create engaging and educational content for her students. One of our goals is to develop custom storybooks and interactive reading materials under the brand Cepat Baca ReadMonster, aimed at encouraging early literacy and making reading fun and meaningful for preschoolers. 

Watching her passionately create learning materials and reading books for her students sparked the idea to build a system that could support her work. So, I decided to create this system for my mom, as a way to support her work with young children in her kindergarten. At the same time, I hope it can also help parents, educators, and children co-create reading materials that feel personal, culturally relevant, and fun to explore — making storytime more meaningful and engaging for everyone.

---

Problem Statement :

• Time Consuming : Traditional methods for creating such materials are often highly time-consuming. It requires a lot of effort and time to find story ideas, develop the story, and eventually publish the book. I saw how my mom often struggled with brainstorming ideas and turning them into meaningful content that is suitable for children.
• Generate Personalized Story : Storybooks are typically created after extensive research, yet they often lack the ability to cater to individual needs. This limitation can result in reduced interest and engagement during storytime. In many kindergartens and homes, parents and teachers rely on a standardized set of storybooks that may not resonate with a child's unique emotions, interests, or developmental requirements.
• Age-Appropriate Stories : The world is becoming increasingly complex, making it harder to find storybooks that are truly suitable for children. I once came across a storybook that appeared to be designed for kids, but its content was actually not appropriate for them.

Can we create a system that generates personalized, age-appropriate stories that reflect values, themes, and characters chosen by the user?

How Gen AI solves the Problem :

To address the challenge of creating engaging, age-appropriate, and personalized storybooks efficiently, I developed a system that integrates multiple Generative AI capabilities. This solution empowers users to craft unique stories based on their preferences and chosen moral values, making storytelling more interactive and meaningful.

Using GenAI evaluation, each story is assessed for age-appropriateness, vocabulary level, and overall quality—ensuring suitability for young readers. Story generation with few-shot prompting allows the system to generate compelling narratives based on just a few examples or prompts, making it easy to create rich and imaginative stories with minimal input. Image generation brings the narrative to life with AI-generated visuals for characters and scenes, allowing a complete storybook to be created in a short time. Meanwhile, grounding tools help users explore and select from trending or meaningful themes, ensuring the storybook is both fun and relevant.  An interactive AI chatbot assistant where you can experience creating personalized stories that cater to your unique needs and preferences. This holistic use of GenAI transforms traditional storytelling into a guided, creative, and educational experience.

We are required to include at least three (3) of these GenAI capabilities in our project. However, I applied six (6) of the capabilities in my project, which are ticked below.

• Structured output/JSON mode/controlled generation ✅
• Few-shot prompting ✅
• Document understanding
• Image understanding ✅
• Video understanding
• Audio understanding
• Function Calling
• Agents ✅
• Long context window
• Context caching
• Gen AI evaluation ✅
• Grounding ✅
• Embeddings
• Retrieval augmented generation (RAG)
• Vector search/vector store/vector database
• MLOps (with GenAI)

Story Generation by GenAI:

Image Generation by GenAI:

Story Evaluation by GenAI:

AI Chatbot using LangGraph Agents :

Limitations and Future Potential of Generative AI in Personalized Storytelling :

Limitations of the Technology :

• Contextual Issues: AI may struggle with maintaining consistency in longer stories.

1. Bias & Inappropriate Content: AI-generated stories might unintentionally contain biases or unsuitable content.

2. Limited Creativity: While AI can generate stories, it might not always produce highly complex or abstract narratives.

3. Image Generation Inconsistencies: AI-generated visuals might not always match the user's exact vision.

4. Overwhelming Customization: Too many options could overwhelm young users in creating personalized stories

Future Potential of Gen AI :

1. Multi-modal Integration: Combining text, images, and audio for richer stories.

1. Improved Coherence: AI can create more consistent and complex narratives.

2. Personalized Learning: Stories tailored to educational objectives.

3. Emotion-Aware Stories: AI generating stories based on the child’s mood.

4. Safety Enhancements: As AI evolves, it will ensure that content is always child-appropriate.

Limitations and Future Potential of Generative AI in Personalized Storytelling

In the future, Generative AI will continue to improve, creating even more engaging and educational personalized stories for children. Despite current limitations, future advancements will make AI even more powerful and intuitive, unlocking its full potential for young readers. However, we have to ensure that we're using AI to improve our lives not to depend on them.

Limitations of Technology:

• Context and Coherence: AI can sometimes struggle with maintaining consistency and logical flow in long or complex narratives, especially with multiple themes and characters.
• Bias and Content Suitability: AI can unintentionally generate biased or inappropriate content, due to limitations in its training data, even with built-in evaluations.
• Creativity Constraints: AI might sometimes go overboard or generate unrealistic story elements since it might struggle in scenarios requiring deeper, nuanced creativity, such as highly complex or abstract storytelling. It can lack the logic and depth needed for more abstract or meaningful storytelling, often needing human input to refine the story.
• Image Generation: AI-generated visuals might not fully align with user expectations, showing inconsistencies in style or detail. The prompt need to be as detail as possible to create visuals that cater to user expectations.
• Customization Overload: Offering too many customizable options (such as themes, character settings, and narrative structures) might overwhelm users, particularly younger ones or those unfamiliar with AI tools.

Future Potential of GenAI:

• Multi-modal Integration: Future iterations could incorporate audio and video understanding, allowing users to create not only stories with text and images but also audio-based stories or even full animated video books. 
• Personalized Learning: GenAI could be used to tailor educational content alongside storytelling. Where AI could adapt the stories to teach specific skills or moral lessons based on the child’s developmental needs, helping them to learn while enjoying the story.
• Emotion-Aware Stories: Leveraging emotion recognition and mood tracking, the system could generate stories based on the emotional state of the reader, make it easy to generate content based on a reader’s mood for therapeutic or uplifting experiences.
• Collaborative Storytelling: Future systems might allow for collaborative storytelling, where children, teachers, or even parents could co-create stories with the AI, using shared prompts and interactions. 
• Adaptive Feedback and Improvements: The integration of ML Ops and feedback loops could lead to continuous improvements in story generation models. User can interact with the system and provide feedback for AI to learn from its past interaction and generate more personalized stories over time.