SignLink is a real-time American Sign Language (ASL) to spoken English translator designed to break communication barriers for the Deaf and Hard of Hearing community.

Built for the Gemini 3 Hackathon 2026, this project leverages the cutting-edge multimodal reasoning of the Gemini 3 Flash model to provide fluid, context-aware translations through a simple web interface.

Most sign language recognition tools rely on complex "hand-tracking" landmarks which fail in low light or at weird angles. SignLink takes a different approach: it uses Gemini 3's native visual intelligence. It "sees" the sign exactly as a human interpreter would, understanding the nuance, speed, and context of the gesture.

This project highlights three core capabilities of the new Gemini 3 ecosystem:

• Multimodal Reasoning: SignLink sends video frames directly to Gemini 3 Flash, which interprets the visual data without needing a secondary "vision" model.
• Low-Latency (Flash): By utilizing the Flash model, we achieve the speed necessary for conversational flow.
• Vibe Coding Logic: The entire frontend and integration were rapidly prototyped using AI Studio, allowing for a focus on user experience rather than boilerplate code.

• Brain: Gemini 3 Flash (via Google Generative AI SDK)
• Frontend: Streamlit (Python-based Web UI)
• Audio: Web Speech API (for Text-to-Speech output)
• Deployment: Streamlit Cloud / Google Cloud

signlink/
├── app.py              # Main Application Logic
├── requirements.txt    # Dependencies
├── .env                # API Key (Local only)
└── README.md           # Project Documentation