VidCoder AI

VidCoder AI
Architecture Diagram

Inspiration

Manually typing code from images is a huge pain. We were inspired to build a smart, all-in-one AI tool that effortlessly extracts, refines, and translates code, leveraging Google ADK concepts and Gemini AI.

What it does

VidCoder AI extracts code from images, then enhances and analyzes it for limitations. It also offers a dedicated tab to translate code between Python, JavaScript, and C++, making it easy to copy or download.

How we built it

We built VidCoder AI as a Streamlit web app, powered by Google's Gemini API (gemini-2.0-flash) and Google ADK. We designed a conceptual multi-agent AI pipeline to handle extraction, enhancement, analysis, and translation.

Challenges we ran into

Accurately extracting diverse code from images and ensuring insightful AI responses were key hurdles. Language detection for short code snippets and orchestrating multiple AI steps within Streamlit also posed challenges.

Accomplishments that we're proud of

We're proud of building a seamless, end-to-end AI-powered workflow for code. We created an intuitive UI and a modular design, effectively solving a real-world problem for developers and students.

What we learned

We gained deep insights into Gemini's multimodal power and got to know about Google's Agent development kit.

What's next for VidCoder AI

Next, we aim to expand VidCoder AI to extract code from PDFs and videos. We'll also add more programming languages and enhance our code review capabilities.