StoryWeaver

Dashboard
Output

Inspiration

Every child deserves a personalized bedtime story. We were inspired by the idea of combining Google's Gemini AI with interactive storytelling to create magical, one-of-a-kind stories for children — complete with AI-generated illustrations that bring each tale to life in real time.

Traditional storybooks are static, but what if a child could choose their own characters, settings, and themes, and watch a brand-new story unfold before their eyes? That's what StoryWeaver does.

What it does

StoryWeaver is an AI-powered interactive storybook that generates personalized children's stories with:

Real-time narrative generation using Gemini 2.0 Flash
AI-generated illustrations interleaved with the story text using native multimodal output
Customizable themes — dragons, space, ocean, fairy tales, robots, pirates, dinosaurs, and more
Age-adaptive content — stories adjust complexity for ages 2–12
Live streaming — story chunks stream over WebSocket for an engaging, page-turning experience
Optional narration via Google Cloud Text-to-Speech

How we built it

Backend: Python FastAPI server running on Cloud Run, using the google-genai SDK to call Gemini 2.0 Flash with interleaved TEXT + IMAGE output modalities. Story chunks stream to the frontend via WebSocket in real time.

Frontend: React + TypeScript with Vite. A split-panel UI with story controls on the left and a rich story player on the right that renders text, images, audio, and status updates as they arrive.

AI Pipeline: We use a two-strategy approach:

Native interleaved output — Gemini generates text and images in a single response
Fallback pipeline — Text-only generation with [IMAGE: ...] tags, followed by Imagen 3 illustration generation

Infrastructure: Dockerized for Cloud Run deployment, with Cloud Storage for generated assets.

Challenges we ran into

Interleaved output parsing — Handling mixed text and inline image data from Gemini's multimodal response required careful part-by-part processing
Streaming UX — Making the story feel like a "page-turning" experience with smooth animations while chunks arrive asynchronously
Age-appropriate content — Crafting system prompts that consistently produce safe, engaging content across different age ranges
Rate limiting — Managing API quotas during development, which led us to build a robust demo mode

Accomplishments that we're proud of

Seamless real-time story streaming with interleaved illustrations
A beautiful, premium dark-themed UI that feels magical and child-friendly
The two-strategy fallback system ensuring stories always generate successfully

What we learned

How to use Gemini's native multimodal output for interleaved text + image generation
WebSocket streaming patterns for real-time content delivery
Building robust fallback pipelines for AI-powered features