AccessiBridge

AccessiBridge is a real-time AI accessibility companion designed to help people with visual and hearing impairments better understand their surroundings using only a smartphone. I built this project to explore how multimodal AI can move beyond productivity and create immediate social impact. The app captures an image, sends it to Gemini for analysis, and delivers clear spoken descriptions of scenes, objects, and text so users can navigate the world more independently.

I built AccessiBridge as a solo developer using React, speech-to-text, text-to-speech, and the Gemini 2.5 Flash multimodal API. The model’s speed and ability to process images, text, and conversational context made real-time assistance possible. During development, I encountered significant API integration challenges, including repeated 404 model and endpoint errors. Debugging model availability, API versioning, and error handling became a key milestone that ultimately enabled reliable scene descriptions and a smoother real-time user experience.

This project demonstrates how fast, affordable multimodal AI can power accessible tools at scale. AccessiBridge shows a future where real-time assistance is always available, affordable, and designed for independence. It represents both a technical achievement and a commitment to building technology that improves everyday life for billions of people.

Built With

• gemini3
• react
• tailwind
• vite