Inspiration
We envisioned a future where human-AI interaction feels natural and immersive. Current chatbots lack personality and visual engagement, so we created AI-NEXUS to bridge this gap. Our inspiration came from combining cutting-edge AI conversation models with lifelike 3D avatars to revolutionize digital communication.
What it does
AI-NEXUS transforms standard chatbot interactions by:
- Giving AI a face and voice through customizable 3D characters
- Generating real-time lip-synced responses using ElevenLabs' voice synthesis
- Providing intelligent conversations powered by OpenAI's language models
- Creating emotional connections through animated facial expressions
How we built it
Tech Stack:
- Frontend: React + Three.js for 3D rendering
- Backend: Node.js with Express
- AI Services: OpenAI GPT + ElevenLabs TTS
- Animation: Rhubarb Lip Sync for mouth movements
- Pipeline: FFmpeg for audio processing
Key Integration Points:
- User input → OpenAI API for text response generation
- Text response → ElevenLabs for voice synthesis
- Audio + text → Rhubarb for lip-sync animation data
- All components synchronized in 3D viewport
Challenges we ran into
- Lip-Sync Timing: Perfecting frame-accurate mouth animations with variable audio lengths
- Latency Optimization: Reducing delay between user input and full character response
- 3D Performance: Maintaining smooth animations while handling complex AI processing
- API Coordination: Synchronizing multiple cloud services (OpenAI → ElevenLabs → Animation)
Accomplishments that we're proud of
✅ Created seamless integration of 3 distinct AI services
✅ Achieved sub-second response times for end-to-end interaction
✅ Developed a reusable animation system for any 3D character rig
✅ Built an extensible architecture for adding new AI features
What we learned
🔧 The importance of audio waveform analysis for natural lip movements
🤖 How to optimize prompts for character-specific AI personalities
⚡ Techniques for parallel processing of AI services to reduce latency
🧩 Integrating multiple npm packages for 3D, audio, and AI workflows
What's next for AI-NEXUS
Near-Term (v2.0):
- Multi-character conversation scenarios
- Emotion detection from text to drive facial expressions
- User-customizable avatar creation
Future Vision:
- VR/AR compatibility for immersive meetings
- AI memory for persistent character personalities
- Real-time video input for responsive eye contact
- Marketplace for community-created characters
Built With
- elevenlabs-api
- express.js
- gltf/glb-format
- node.js
- openai-api
- react-three-fiber
- rest
- rhubarb
- tailwind
- three.js
- typescript
- vite
- web-audio-api
- websocketreact


Log in or sign up for Devpost to join the conversation.