Llama-Cook

Your lightning-fast AI cooking companion, delivering instant guidance and predictive assistance through every step.

Lightning-Fast Response

Llama-4-Scout delivers instant answers with 1M token context

Real-time Visual Recognition

SAM2 + Llama-4-Scout identify ingredients instantly

Low-Latency Voice Control

Scout's efficiency enables seamless hands-free cooking

Watch Sous Chef in Action

Creating fresh, healthy meals

Meta Llama APIAI
Meta SAM2Vision
Meta V-JEPA 2Prediction
AWS TranscribeSpeech-to-Text
AWS PollyText-to-Speech
BoundaryMLStructure
Meta Llama APIAI
Meta SAM2Vision
Meta V-JEPA 2Prediction
AWS TranscribeSpeech-to-Text
AWS PollyText-to-Speech
BoundaryMLStructure
Step 1: Voice Command

"What can I do with these ingredients?"

Start your cooking journey with a simple voice command

AWS Transcription

Demo

"Sous Chef, what can I do with these ingredients?"

Wake word detected

iPhone Camera Stream

Streaming

Your Hands-Free Kitchen Journey Begins Here

With just your voice, unlock personalized recipes based on what's in your kitchen. No typing, no scrolling – just natural conversation with your AI sous chef.

1Wake Word Detection

Simply say "Sous Chef" to activate your assistant. Powered by AWS Transcribe's real-time streaming, it's always ready when you are.

2Natural Language Understanding

Ask naturally: "What can I make with tomatoes and mozzarella?" The system understands context and cooking intent.

3Instant Voice Response

Get immediate, conversational responses through Amazon Polly's neural voices. Sub-100ms latency keeps the conversation flowing naturally.

Try it yourself:

"Sous Chef, I have fresh tomatoes, mozzarella, and basil. What can we make?"

→ "Perfect! With those ingredients, we can make a classic Caprese Salad. Would you like me to guide you through it?"

Step 2: Send to Llama API

AI-Powered Recipe Generation

Contextual Understanding: Llama-4-Scout analyzes your ingredients and dietary preferences

Instant Suggestions: Get personalized recipe ideas in under 100ms

Step-by-Step Guidance: Conversational cooking instructions tailored to your skill level

Adaptive Learning: Remembers your preferences for future recommendations

Structured Output: Powered by BoundaryML for consistent, well-formatted recipe suggestions

Voice Synthesis: AWS Polly neural voices bring Sous Chef to life with natural, expressive speech

Llama-4-Scout-17B Response

Processing
Analyzing ingredients
Processing...
Step 3: Vision Tracking

SAM2 keeps watch of ingredients and their state

Real-time tracking ensures nothing is overlooked during preparation

Basil - Fresh
Tomato - Whole
Mozzarella - Whole
Basil
Tracked
Tomato
Ready
Mozzarella
Waiting

Your AI Kitchen Observer

SAM2's advanced segmentation technology watches your ingredients throughout the cooking process, tracking their location and state changes to ensure perfect timing and coordination.

1Automatic Detection

SAM2 identifies and tracks each ingredient without manual labeling, understanding what needs attention.

2State Recognition

Monitors changes like chopping, mixing, or cooking, alerting you when ingredients need attention.

3Position Tracking

Keeps track of where everything is, even if temporarily obscured or moved around the kitchen.

SAM2 Technology: Zero-shot segmentation • Real-time tracking • 51K+ videos trained

Step 4: Prediction AI

V-JEPA 2 anticipates your next move

Keeps your recipe on track by predicting and guiding each step

Your Predictive Cooking Assistant

V-JEPA 2's world model watches your cooking progress and anticipates what comes next, providing timely reminders and preventing common mistakes before they happen.

Anticipating Your Actions

"I see you're about to slice the tomatoes. Remember to use a sharp knife and cut them 1/4 inch thick for even layering."

Predicted next: Tomato slicing in ~30 seconds

Real-Time Guidance

Currently: Arranging mozzarella slices

Next: Layer tomato slices

Then: Add fresh basil leaves

Finally: Drizzle olive oil

Mistake Prevention

!

Timing Alert

Don't forget to season between layers

Technique Check

Good knife angle for clean cuts

V-JEPA 2 Technology: World model • 62+ hours training • Motion prediction

Motion Prediction Timeline

30s ago

Washed basil leaves

Now

Slicing mozzarella

In 45s

Start tomato slicing

Confidence: 92%

In 2m

Layer ingredients

Confidence: 87%

In 3m

Add olive oil drizzle

Confidence: 95%

Model Confidence94%
Step 5: Experience It All

See the complete AI cooking experience

Watch all the technologies work together in real-time

Ready to Cook with AI?

Quick Start Guide

1

Set up your iPhone camera streaming

2

Place ingredients on your counter

3

Say "Sous Chef" to activate voice

4

Follow AI guidance to create your dish

All Technologies Working Together

Voice AI
Active
Llama-4-Scout
Processing
SAM2 Vision
Tracking
V-JEPA 2
Predicting
47ms

Response Time

98%

Accuracy

Demo Requirements

iPhone with camera + MacBook Pro + Local server running

Setup Instructions

Get the demo running in minutes

Prerequisites

Before starting, ensure you have:
• Node.js 18+ installed
• Python 3.10+ with pip
• CUDA-capable GPU (for SAM2)
• iPhone and MacBook on same network
• Meta AI API keys

Quick Start Tip

For the hackathon demo, we recommend running all services locally on the MacBook for best performance. The setup can be completed in under 10 minutes with all dependencies installed.