Inspiration

As a huge foodie, I started cooking more at home to recreate the meals I couldn’t always go out for. But trying to manage a sizzling pan, sharp knives, and a phone at the same time was bound to end badly. That frustration sparked the idea for ChopChop—a hands-free, AI-powered voice cooking assistant that lets you navigate recipes, get guidance, and keep cooking safely using only your voice.

What it does

ChopChop is an AI-powered, hands-free voice cooking assistant that acts as your personal kitchen companion. Pick any recipe, and ChopChop walks you through each ingredient to get your mise en place ready (yes, the puns continue), then guides you step-by-step through the cooking process using natural voice commands.

  • Missed an instruction? Have it say it again.
  • Unsure about something you already did? Tell it to rewind.
  • Know the recipe inside out? Fast-forward to any step—or straight to the finish.
  • Messed it all up and want a fresh start? Just say the word. All of this happens through real-time voice interaction, with live speech recognition, an animated orb that reacts to your voice, and smart sleep/wake behavior so ChopChop is always ready when you are—without getting in the way.
  • Hands-Free Recipe Navigation - Voice-controlled step-by-step cooking guidance
  • Natural Conversation - Ask questions, repeat steps, skip ahead, or go back
  • Real-Time Responses - Powered by Deepgram's Agent API with minimal latency
  • Visual Feedback - Animated orb responds to your voice
  • Smart Sleep Mode - Auto-sleeps after 30 seconds, tap to wake
  • Multiple Voices - Choose from 4 different AI voice personalities
  • Smart Grocery Lists - Auto-generate shopping lists from recipes
  • Cost Savings - Get cheaper ingredient substitution suggestions
  • Shopping Integration - Send ingredients to Instacart, Walmart, or Amazon
  • YouTube Integration - Ask to watch recipe videos on YouTube
  • Maps Integration - Find grocery stores near your location

How we built it

We built ChopChop using Next.js with React for the frontend, leveraging Deepgram's Voice Agent API for real-time voice conversation capabilities. The architecture uses WebSocket connections for bidirectional audio streaming, the Web Audio API for microphone capture and playback, and GPT-4o as the conversational AI brain. We implemented custom audio processing utilities to handle sample rate conversion (16kHz input, 24kHz output) and created context providers to manage microphone, WebSocket, and conversation state across the app. We used Kilo code to deploy the app to vercel.

Challenges we ran into

Managing real-time audio synchronization was complex. We had to buffer and schedule playback precisely to avoid gaps or overlaps. Handling interruptions (when users speak while the AI is talking) required careful state management to clear audio buffers and reset conversation flow. Mobile browser limitations around wake locks and audio contexts also demanded platform-specific workarounds. Balancing low latency with clear, natural audio was an ongoing challenge.

Accomplishments that we're proud of

We were able to build ChopChop using Deepgram's Voice Agent API that serves real time voice and can have conversation to assist the users.

What we learned

We learnt to use Deepgram's Voice Agent API for Real-time audio processing in web browsers

What's next for ChopChop

Next, we plan to add voice-based recipe search and database integration, smart timers ChopChop can manage for you, and multi-language support. We’re exploring ingredient substitutions, dietary awareness, and a personalized “learn mode” where ChopChop adapts its guidance style to your cooking skill and preferences.

Built With

Share this project:

Updates