HeyGen Hackathon · Agent Track
A customer in Spain gets a damaged order. They write in Spanish — angry, frustrated. The support bot replies in English. They don't understand. They leave. Forever.
This happens 2.5 billion times a day. Language barriers cost businesses $50B+ annually in lost customers.
Customer types in ANY language → AI understands → Avatar SPEAKS BACK in their language. Lip-synced. On video.
"¡Mi pedido llegó dañado!" (Spanish)
↓
Agent detects: Spanish · Angry · Complaint
↓
LLaMA writes empathetic response in Spanish
↓
HeyGen Avatar speaks it — perfectly lip-synced
↓
Customer feels heard. Problem solved. In 10 seconds.
No human agent needed for 80% of tickets.
Video is not a chatbot. When a customer is furious, a text bubble doesn't calm them down. A human face speaking their language does. That's the entire insight.
- 🌍 Works in Spanish, Hindi, French, German, Arabic, Japanese and more
- 🎭 Real avatar. Real lipsync. Not dubbed — generated natively in their language
- ⚡ 10 seconds from complaint to video response
- 🤖 Fully autonomous — no human in the loop
| Who benefits | How |
|---|---|
| E-commerce brands | Stop losing non-English customers |
| Healthcare | Patients understand discharge instructions in their language |
| Banks & Fintech | Serve emerging markets without hiring multilingual staff |
| Any global business | 24/7 support in every language, zero extra headcount |
The global customer service market is $500B. Every company with international customers needs this.
React UI → FastAPI → LangGraph Agent → Groq LLaMA → HeyGen Avatar
5 agent nodes, fully autonomous:
- Detect — Groq LLaMA identifies language + sentiment + intent
- Route — Non-English or angry? → Avatar. Simple English? → Text
- Script — Writes empathetic response in customer's native language
- Generate — HeyGen Avatar V renders speaking video with lipsync
- Deliver — Video streams to customer in real time
HeyGen tools used: Avatar V · Lipsync · Video Translate · Hyperframes · Video Agent
| Layer | Technology |
|---|---|
| Frontend | React + Vite |
| Backend | FastAPI + Python |
| Agent | LangGraph + LangChain |
| LLM | Groq LLaMA 3.3 70B |
| Avatar + Video | HeyGen API |
# 1. Add your keys
cp backend/.env.example backend/.env
# → Add GROQ_API_KEY and HEYGEN_API_KEY
# 2. Backend
cd backend && pip install -r requirements.txt
uvicorn main:app --reload --port 8000
# 3. Frontend
cd frontend && npm install && npm run dev
# 4. Open http://localhost:3000
# Click 🇪🇸 ES → watch the avatar speak SpanishYou need 2 keys. That's it.
- Groq → console.groq.com (free)
- HeyGen → app.heygen.com (free trial)
Click the Spanish button. Watch the agent pipeline fire live on screen. 10 seconds later — an avatar looks at the camera and speaks Spanish. Lip-synced. Empathetic. Perfect.
That's the moment that wins the room.
Built in 24 hours · HeyGen Hackathon 2026