Inspiration
After watching xQc roast a sponsorship live on stream, we realized how much a streamer’s tone can make or break a brand. That moment sparked Streamsense: a way to measure live sentiment and brand perception in real time.
What it does
Streamsense tracks mentions of sponsored products on stream and analyzes both streamer sentiment and chat reactions. It visualizes real-time audience response to help sponsors understand how their campaigns actually land.
How we built it Built with Python and Streamlit, Streamsense uses CLIP embeddings for finding brand images within livestreams, Twitch API for data ingestion, and WebSockets for live chat tracking. The app displays dynamic dashboards showing sentiment over time and key mention moments.
Challenges
APIs broke mid-demo, WebSockets dropped constantly, and aligning streamer speech with chat timing was pure chaos, but we learned a lot about debugging real-time systems under pressure.
Accomplishments
We got a live prototype working that successfully captured sentiment spikes from chat reactions and streamer tone—plus we fixed the random crashes (mostly).
What’s next
We’re planning to improve accuracy with multimodal sentiment models, expand to YouTube and Kick, and harden the backend for scalability and reliability.
Log in or sign up for Devpost to join the conversation.