Teddy.AI

• 

  Pictures are AI generated to go along with your story!

• 

Inspiration

My cousin recently had children, and as they enter their toddler years, I saw how they use their toys. They buy them, use them for a couple weeks, but then get bored. To be honest, however, I can't blame them. Currently, toys are used in one way, with no real interaction coming from both the child and the toy. What I really wanted to do was bring Toy Story to real life, and allow children to talk and learn from their toys, maximizing their happiness and education.

What it does

We use Hume's API to allow you to talk with your toys and have full blown conversations with them. We utilize prompt engineering and allow you to have math lessons embedded within their choose your own adventure stories.

How we built it

We embedded a raspberry pi, speaker, and microphone in the animal, which hosts a web app through which it can speak.

Challenges we ran into

The Hume API was down sometimes which was tough to navigate, which halted our ability to make a self improving prompt (track progress of kids lessons). We also had a broken raspberry pi for the first 12 hours of the Hackathon. Hyperbolic randomly stopped working for us so our interactive story pictures stopped working.

Accomplishments that we're proud of

Learning about how to prompt image generation (feeding the text transcription into Hyperbolic)

What we learned

Image generation, prompt engineering, function calling

What's next for Teddy.AI

Memory and lesson progress reports

Built With

• gemini
• generativeai
• hume
• hyperbolic
• javascript
• large-language-models
• node.js
• raspberry-pi