Storyteller

StoryTeller is an experimental web application that creates short audio stories for pre-school kids.

Example Story

demo-recording.mp4

How it works

The main flow from the diagram can be found in generateStoryFlow. Most of the UI is implemented in index.tsx.

Development

1. Install dependencies: npm i

2. Setup .env with required keys and base url:

    NEXT_PUBLIC_BASE_URL="http://localhost:3001"
    OPENAI_API_KEY="sk-..."
    ELEVENLABS_API_KEY="..."
    LMNT_API_KEY="..."
    STABILITY_API_KEY="sk-..."
   

   You can get the API keys from the respective services:

   • OpenAI
   • Eleven Labs
   • Lmnt
   • Stability

3. Start Fastify server (runs on port 3001): npm run fastify

4. Start Next.js server for UI (runs on port 3000): npm run next

5. Go to localhost:3000

Deployment

1. Build the Next.js app: npm run build. This exports the app to the out folder, so it can be served by Fastify.

Deploy to Render

StoryTeller can be deployed as a Render web service.

• Build command: npm i; npm run build
• Start command: npm start
• Configure the NEXT_PUBLIC_BASE_URL environment variable to the Render service URL (base url must not have trailing slash)
• Add the API keys to the Render Environment settings.

Used Libraries and AI services

StoryTeller is built with the following libraries:

• ModelFusion
• Fastify
• Next.js
• shadcn/ui
• Zod

The following AI APIs are used:

• OpenAI
• Eleven Labs
• Lmnt
• Stability

Limitations

• No persistent execution with error handling
• No user accounts / persistent data (only storage of image and audio files)
• No user authentication / authorization
• Limited error handling on the client
• Tested on Chrome Desktop (Mac). Other browsers & operating systems may not work.