ZeraCam

ZeraCam is a cutting edge tool that bridges the gap between the physical and digital worlds, allowing AI to “see” code on a screen, detect errors, and suggest fixes in real-time. I was inspired to build this project after noticing how frustrating debugging can be sometimes errors are obvious on the screen, but understanding and fixing them takes time. I wanted to create an AI assistant that doesn’t just read code but truly understands it, highlights issues visually, and explains why the errors occur. Building ZeraCam taught me a lot about combining computer vision with code analysis, handling real-time webcam streaming, and creating smooth communication between a Python backend and a React frontend. One of the most interesting parts was integrating Gemini 3’s AI to analyze code frames, provide corrections, and even overlay them on the screen with bounding boxes. For example, it can recognize a missing operator in Python and suggest the fix instantly, like turning result = a b into result = a + b. The project wasn’t without challenges lighting variations, real-time latency, recognizing handwritten code, and accurately mapping AI suggestions to the screen all required careful solutions. Despite these hurdles, creating ZeraCam was an incredible learning experience, showing me how AI can act as a proactive assistant that understands, explains, and helps improve code in real-time. I’m proud of how it turned into a tool that feels almost like a “Code Ghost,” silently watching over your shoulder and making programming smoother and more intuitive.

Built With

• opencv
• pillow
• python
• typescript