Re-Bias is an AI-powered web and mobile system that reveals how biased social media content truly is — across YouTube, TikTok, and Instagram Reels.
Instead of trying to remove bias, Re-Bias visualizes it transparently, empowering users to understand and rebalance their media perspective.
Modern recommendation algorithms often amplify political or emotional bias.
Re-Bias quantifies these tendencies by analyzing the language, tone, and visual mood of each video — returning a Bias Index (0–100%) with interpretive reasoning.
The system is designed to help users become aware of subtle influence rather than filter it away.
- Extracts transcripts, audio sentiment, and visual cues from uploaded or linked videos.
- Detects political, ideological, and emotional bias using AI models.
- Generates a unified Bias Index and reasoning summary.
- Displays a live bias banner overlay on the video feed (similar to a phone battery icon).
- Color-coded indicator:
- 🟢 Neutral (0–30%)
- 🟡 Moderate (31–60%)
- 🔴 High (61–100%)
- Tapping the banner opens the full in-app analysis page.
- Helps users refresh their YouTube/TikTok algorithm feed to reduce echo-chamber effects.
| Layer | Technology | Purpose |
|---|---|---|
| Frontend | HTML, CSS, JavaScript | UI, transcript extraction, bias visualization |
| Backend (Planned) | Python (Flask / FastAPI) | API layer, AI model integration |
| AI Models | Whisper / BERT / RoBERTa / OpenCV | Speech-to-text, sentiment & bias detection, scene analysis |
| Tools | Node.js, GitHub, VS Code | Development environment & collaboration |
| Hosting (Future) | Vercel / AWS | Deployment and scalability |
- User uploads or links a video (YouTube, TikTok, Instagram Reels).
- System extracts text and audio data.
- AI models analyze sentiment polarity and topic framing.
- The Bias Index is calculated.
- The result is visualized directly on the video feed or via the app dashboard.
🎬 Demo on SNS Platforms (as a Banner): Watch on YouTube
📱 Demo of our own app: Watch on YouTube