StreamWise

Platform interface
Video upload and preview
Load analysis
Analysis dashboard
Language analysis breakdown
Emotion analysis breakdown

Inspiration

With the rapid rise of eCommerce live streaming, many viewers experience frustration with low-quality, unengaging broadcasts. We wanted to create a solution that not only helps viewers find high-quality content but also empowers creators to improve. Our goal was to build a platform that quantifies live stream quality, promotes transparency, and enhances the user experience, ultimately contributing to a better reward system for creators.

What it does

StreamWise quantifies the quality of eCommerce live streams, providing an objective score based on key factors like energy, pacing, audience interaction, and content clarity. This helps viewers identify engaging content and empowers creators to improve their performance by highlighting areas for growth. By offering a measurable evaluation of live streams, it supports the creation of a fairer, more transparent reward system.

How we built it

We used a combination of Streamlit for the interactive frontend, faster-whisper for speech-to-text analysis, and OpenCV for video quality checks. We also integrated MediaPipe for emotion analysis to assess the presenter's engagement and Tesseract OCR for text extraction in the live stream. Data processing and analysis were done with Python, leveraging libraries like Pandas and NumPy.

Challenges we ran into

Unfamiliarity with Video Analysis: Working with video content was a completely new area, and it took significant time to explore and learn the tools and techniques needed for accurate analysis.
Finding Quality Test Videos: It was difficult to find diverse test videos that effectively demonstrated differences in content quality, making it harder to fine-tune the system.

Accomplishments that we're proud of

Team Collaboration: Despite the challenges of working during the semester, our team successfully maintained clear communication, aligned on key details early, and kept an open line of discussion throughout the process.
Working Web App: We successfully built a functioning web app that enables video analysis, contributing to the reimagination of a more transparent and effective reward system for eCommerce live streaming.
Deeper Understanding: We gained a much clearer understanding of the key elements that define high-quality eCommerce live streams, including audience engagement, energy levels, and presentation clarity.

What we learned

Critical Elements of a High-Quality Stream: Through our analysis, we learned the importance of factors like presenter energy, pacing, and audience interaction in determining the success of a live stream.
Dynamic of Live Content: We now have a deeper understanding of how live stream content needs to be dynamic and responsive, rather than static, to keep audiences engaged.
Tech Integration: Integrating various analysis tools like speech-to-text, emotion recognition, and video quality assessment required us to better understand how these components can work together to create a seamless user experience.

What's next for StreamWise

Faster Processing for Longer Videos: Currently, the system can only process shorter videos, and longer videos take time to analyze. Improving the processing time for longer content is a priority.
Enhanced Visualizations: We aim to improve the way each quality criterion is visualized, providing users with clearer, more actionable insights into the strengths and weaknesses of their streams.
AI-Driven Recommendations: Moving beyond simple scores, we want to build AI-driven recommendations that offer personalized advice to creators on how to improve specific aspects of their live streams.