1-2-Tree

Login Page
Sign up Page
Music Mode
Ai Generated Song
Card Mode
Ai Generated Flash Cards
Song Library
Flash Card Library

Inspiration

We were inspired to build 1-2-Tree because many children around the world lack reliable access to high-quality preschool education. Early childhood (ages 3–5) is when kids build the foundations of language, numbers, and emotional skills, but access often depends on whether families can afford daycare, find open spots nearby, or create their own learning materials at home. At the same time, many households do have a phone, tablet, or laptop. We saw an opportunity to turn that everyday device into a playful but structured learning companion for families and teachers, using AI to generate age-appropriate rhymes and learning content instead of constantly asking adults to invent new activities from scratch.

What it does

1-2-Tree is a web application that turns any preschool topic into simple, educational content. The User types in a word or theme (like “colors”, “sharing”, or “butterflies”), and the system uses AI to generate gentle, age-appropriate rhymes and song-style learning material that can be used as cards, activities, or story time content. The experience is designed to feel like a fun game for kids while still giving parents and teachers a structured tool they can reuse in lessons

How we built it

On the frontend, we built a React 18 + TypeScript app using Vite, React Router v6, and Tailwind CSS for a fast, responsive, and kid-friendly UI. On the backend, we use Python with FastAPI, Celery, and Redis to handle asynchronous generation jobs, and MongoDB Atlas to store users, their libraries, and cached AI outputs. Gemini AI powers the educational text/lyrics generation, while tools like ElevenLabs and Bark handle high-quality audio where needed. Everything is containerized with Docker and can be deployed via services like Render or similar platforms.

Challenges we ran into

We had to stitch together several moving pieces—Gemini for content, ElevenLabs/Bark for vocals, custom background tracks, and FastAPI + Celery + Redis for async processing—while keeping the overall generation time reasonable. Getting the audio pipeline right (text → vocals → background music → mixing → timing) was especially tricky; small mismatches could break the experience. We also had to design a secure, multi-user system with authentication, per-user libraries, and caching on a tight hackathon timeline, which forced us to be very disciplined about schema design and API boundaries

Accomplishments that we're proud of

We shipped a fully working end-to-end experience: a polished React + TypeScript frontend with a friendly interface and personal library, backed by a FastAPI service that can generate complete educational content from a single topic. On the backend, we built an asynchronous pipeline with Celery and Redis that combines AI-generated text, vocals, and background music into finalized outputs stored in MongoDB for instant replay. We’re also proud that we implemented real security practices—password hashing, validation, and environment-based configuration—instead of cutting corners for the demo.

What we learned

We learned how to design and operate a small but realistic distributed system: a React frontend talking to a FastAPI backend, which in turn coordinates Celery workers, a Redis broker, and MongoDB Atlas. We deepened our understanding of prompt engineering with Gemini 2.5 Flash to consistently get safe, age-appropriate content, and how to tune ElevenLabs so it produces natural, singable speech instead of flat TTS. We also gained a much better appreciation for audio tooling and for how important UX decisions are when you’re building for very young learners and busy parents.

What's next for 1-2-Tree

Next, we want to layer more structured learning on top of the generated content: printable/interactive cards generated from each topic, simple quizzes, and themed packs (numbers, colors, emotions, nature, etc.). Using the same backend, we can add multi-language support and difficulty levels so content grows with each child. On the platform side, we plan to refine performance and deployment and run small pilots with parents and preschool teachers to validate educational impact before scaling.