Inspiration

As of 2020, the foreign-born population makes up about 12% of the total U.S. population. These second-generation immigrants grow up with the unique struggle of balancing two cultures and reconciling the clashes between them.

As children of immigrants ourselves, we know what it feels like to think and speak in a language different from our parents and relatives. This is why we created StoryBridge—to help children connect with their heritage while embracing their new language.

What It Does

StoryBridge is an AI-generated story creation platform that empowers parents, teachers, and young students to create personalized bilingual storybooks in English and a second language.

From original adventures to classic fairy tales, StoryBridge helps children bridge the gap between their culture and the culture of the country they were born in, making language learning a natural and engaging experience.

How We Built It

We gathered feedback from hackers on their language-learning experiences to shape our storytelling approach.

  • Using OpenAI’s GPT-3.5-Turbo, we generate bilingual stories with few-shot prompting and context retention, ensuring natural language acquisition.
  • For visuals, GPT-3.5-Turbo creates image descriptions, which Luma Labs transforms into high-quality illustrations. A polling system with retry logic manages asynchronous image generation.
  • To optimize performance, we implemented state management for cached images, page navigation, and loading states, along with a parsing engine for structured data extraction.
  • Our prompt engineering system ensures multi-language support with relevant vocabulary selection.

Built with Next.js, our full-stack framework allows seamless integration of AI-driven storytelling, providing a responsive and interactive user experience.

Challenges We Ran Into

One of the biggest challenges was ensuring that the AI-generated images maintained a consistent aesthetic throughout the story. We had to fine-tune prompts and metadata to ensure the illustrations had a uniform art style and character consistency across pages.

We also had to carefully curate translated words within the stories. It was important that vocabulary was useful and practical while ensuring children could understand new words through context and visuals, rather than just direct translations.

Balancing image consistency, meaningful word selection, and seamless storytelling required multiple iterations, but refining these elements made our platform more effective for language learning.

Customer Validation

Our team is bilingual, and many of us have firsthand experience navigating the challenges of language retention and heritage preservation. Additionally, we actively teach bilingual and English language learners (ELLs), giving us direct insight into the struggles children face when learning multiple languages.

To ensure StoryBridge effectively meets the needs of young learners, we conducted extensive outreach:

  • We spoke with fellow hackers about their experiences growing up bilingual, discussing how language and heritage intertwine, the difficulties of language retention, and the challenges they faced.
  • We consulted elementary school teachers to understand the obstacles children encounter when learning a new language and to gain their perspectives on effective bilingual education strategies.
  • We analyzed how storytelling can reinforce both language learning and cultural identity, ensuring that the words and context provided in our stories align with real-world language learning needs.

Through this user-driven validation process, we tailored StoryBridge to create an experience that is not only engaging but also practical and effective in helping bilingual children retain and embrace their languages.

Accomplishments That We're Proud Of

We’re incredibly proud to have built a fully functioning system that dynamically generates personalized bilingual stories.

  • We ensured that the words being taught are relevant and practical for everyday use.
  • Through user testing, we verified that children could understand the stories and actually learn meaningful vocabulary through context and imagery.
  • We achieved character consistency across AI-generated images, creating a cohesive visual storytelling experience.
  • Our LLM hallucination rate was extremely low, ensuring accurate translations, relevant vocabulary, and logical storytelling.

These milestones demonstrate the potential of AI-driven bilingual education, and we’re excited to keep improving!

What We Learned

We learned how to integrate OpenAI’s GPT-3.5-Turbo for story generation and Luma Labs for image creation.

Additionally, we gained experience in:

  • Few-shot prompting and context retrieval
  • Caching and prompt engineering to improve accuracy and consistency
  • Front-end development and child-friendly UX/UI design to create an engaging experience

This project deepened our understanding of AI-driven education and its potential for enhancing language learning.

What's Next for StoryBridge

We plan to take StoryBridge beyond static storytelling by using Luma Labs to create animated videos, where it feels like a family member is telling the story with multi-language voice capabilities.

We also want to introduce conversational storytelling, allowing children to interrupt at any time to ask questions, making learning more interactive and engaging.

Additionally, we aim to make stories downloadable as EPUB files, so families can access them on the go, even without an internet connection.

By enhancing interactivity, accessibility, and engagement, we’re working to make StoryBridge an even more powerful tool for bilingual learning.

Built With

Share this project:

Updates