Inspiration:
Most new streamers start with <5 viewers, making it impossible to practice audience interaction and build confidence. We wanted to create a realistic streaming experience that helps streamers develop their skills without the pressure of real viewers.
What it does:
Sup, Chat? is an AI-powered streaming confidence builder that creates realistic fake audience interactions through face recognition, emotion detection, and intelligent AI chat personalities based on real influencers. It simulates a complete streaming experience with donations, viewer counts, and contextual AI responses to help new streamers practice and build confidence.
How we built it:
Built with React frontend and Node.js backend using Socket.IO for real-time communication. Integrated OpenAI's GPT-4o-mini for AI personalities, face-api.js for computer vision, and Web APIs for camera/microphone access. Implemented periodic face detection, emotion recognition with throttling, and a complete donation system with sound effects and text-to-speech.
Challenges we ran into:
- Face Detection Performance: Continuous detection caused 60%+ CPU usage. Solution: Periodic detection every 3 seconds with confidence thresholds.
- AI Response Quality: Early responses felt robotic. Solution: Fine-tuned personality prompts with real streamer personas.
- Real-time Communication: Managing multiple Socket.IO events led to memory leaks. Solution: Comprehensive cleanup functions and event throttling.
- Cross-browser Compatibility: Different browser implementations for speech recognition and WebRTC. Solution: Feature detection and fallback mechanisms.
Accomplishments that we're proud of:
Created the first AI-powered streaming confidence builder that combines real-time face recognition, emotion detection, and intelligent chat personalities to help new streamers practice with realistic audience interactions.
What we learned:
- Technical Skills: Mastered computer vision, AI integration, and real-time communication.
- User Experience: Learned the psychology of streaming and importance of authentic interactions.
- Problem-Solving: Developed strategies for cross-browser compatibility and performance optimization. - AI Integration: Gained expertise in prompt engineering and personality-based response generation.
What's next for Sup, Chat?
Enhanced AI: Expand to 50+ personalities with custom creation tools. Advanced Emotions: Add anger, surprise, fear detection with intensity levels. Multi-language: Support Spanish, French, German, Japanese streaming. Analytics: Track improvement over time with detailed metrics. Mobile App: Native iOS/Android apps for on-the-go practice. Cloud Integration: Store face training data and sync preferences across devices.
Built With
- css3
- express.js
- face-api.js
- html5
- javascript
- node.js
- openaiapi
- react
- socket.io
- tensorflow.js
- webaudioapi
- webrtc
- websocketapi
- webspeechapi

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