Noctale

System Design!

🦉 Inspiration

Bedtime stories are a cherished part of childhood, but kids rarely get to be the creators. We wanted to build something that turns their imagination into stories they can see, hear, and feel proud of.

✨ What it does

Noctale transforms children’s drawings into magical illustrated storybooks. With just an image upload, our AI recognizes the characters, builds a narrative around them, and generates cozy bedtime stories, making every child both the artist and the author. In addition, the reading doubles as an educational experience, allowing parents to engage children in conversational reading comprehension instead of rigid testing. Research on dialogic reading shows that open-ended discussions help children develop deeper understanding and stronger language skills than traditional quizzes. Conversations let kids express reasoning, connect ideas, and build narrative skills, turning storytime into a warm and reflective exchange that nurtures both imagination and cognitive growth.

🛠️ How we built it

We built a full-stack web application using Next.js, FastAPI, and MongoDB, integrating Google Vertex AI (Gemini) for image interpretation and story generation. Drawings are uploaded to Google Cloud Storage buckets, where our backend processes them into illustrated story pages. To maintain visual consistency, our image pipeline references previous frames and character outlines to preserve details across scenes. We incorporated a knowledge base aligned with state reading standards to generate adaptive comprehension prompts that support early literacy development. Voice interactions are powered by ElevenLabs for natural narration and Gemini Speech-to-Text, with Document AI analyzing children’s spoken responses to provide insight into understanding. Together, these systems create an immersive storytelling experience that blends creativity with personalized learning.

⚙️ Challenges we ran into

Maintaining character consistency between pages was one of our biggest challenges. We experimented with various prompt-tuning methods before discovering that referencing the previously generated image as contextual input produced the most coherent results. This approach mimics how AI models learn from sequential patterns: Using prior visual cues to guide the next generation step. We also faced difficulties aligning different image formats between editing APIs and optimizing latency during story generation, balancing image quality with real-time responsiveness.

🏆 Accomplishments that we're proud of

We created an experience that feels magical for kids and meaningful for parents. The storytelling is cozy, character-driven, and emotionally engaging, while the interface remains simple and accessible for anyone to use. Beyond creativity, each story becomes a gentle learning moment. Parents can guide open-ended conversations and comprehension exercises that help children think critically about what they read, turning storytime into both an imaginative and educational experience.

📚 What we learned

We learned how to combine generative AI models for image and text creation, refine prompt chaining for consistency, and manage secure integrations within the Google Cloud ecosystem. More importantly, we discovered the joy of building technology that inspires creativity and brings families together through storytelling.

🚀 What's next for Noctale

We plan to add real-time feedback and multi-page story arcs, along with classroom sharing features that let teachers turn art projects into collaborative storybooks. With these additions, Noctale will evolve into a digital library where every child’s imagination can grow, and little hands can continue making big stories.