🎤 Next.js Voice AI Assistant on Cloudflare

A real-time voice AI assistant built with Next.js and deployed on Cloudflare Workers. Features voice activity detection, speech-to-text, AI conversation, and text-to-speech with WebSocket communication and Durable Objects.

✨ Features

• 🎙️ Real-time voice conversation with AI
• 🔍 Voice Activity Detection (VAD) for automatic speech detection
• 🗣️ Speech-to-Text and Text-to-Speech integration
• ⚡ WebSocket communication for real-time audio streaming
• 🌐 Cloudflare Durable Objects for stateful voice sessions
• 🚀 Edge deployment for global performance

🚀 Quick Start

1. Generate types:

   cd worker && yarn cf-typegen

2. Start the Durable Object Worker:

   yarn dev

3. Set up environment:

   cp .env.example .env.local

4. Start Next.js development server:

   yarn dev

5. Open http://localhost:3000 and start talking!

Note: Both the Durable Object Worker and Next.js server must be running simultaneously.

🚀 Deploy to Cloudflare

1. Login to Cloudflare:

   npx wrangler login

2. Deploy the Durable Object Worker:

   cd worker && yarn deploy

3. Update environment variables:

   • Copy the Worker URL host (e.g., worker-unique-identifier.workers.dev)
   • Set NEXT_PUBLIC_WS_HOST in .env.local to this value

4. Deploy Next.js app:

   yarn deploy

Alternative: Connect your repository for automatic deployments on every push.

🛠️ Tech Stack

• Frontend: Next.js 15, React 19, TypeScript, Tailwind CSS
• Backend: Cloudflare Workers, Durable Objects
• Voice: Voice Activity Detection (VAD), WebSocket streaming
• AI: Cloudflare AI, Speech-to-Text, Text-to-Speech
• Deployment: OpenNext.js for Cloudflare

📚 References

• talk2ai - Cloudflare AI Voice Assistant
• Build Live Cursors with Next.js, RPC, and Durable Objects