-
-
KinoPro Logo
-
Thumbnail
-
Generated Poster
About KinoPro
KinoPro is my submission for the online Devpost Gemini 3 Hackathon. It helps creators turn long-form video into trailer-ready assets by generating storyboard clips, emotional beats, music cues, poster candidates, and export-ready outputs (Video, EDL, XML, JSON).
What inspired me?
I was inspired by how painful trailer pre-production still is. Editors and creators spend hours scrubbing footage, picking moments, and manually building a narrative arc before they can even start cutting. I wanted to build a tool that gives teams a fast “first creative pass” so they can focus on taste and storytelling instead of repetitive prep work.
How I built it?
I built KinoPro as a full-stack web app with a React/Vite frontend and a Python backend service pipeline. Gemini 3 Flash powers the video understanding and structured storyboard generation. I combined AI outputs with media processing to produce clip thumbnails, timestamps, beat labels, compare views across multiple storyboard variants, and one-click exports for editing workflows. I containerized the stack for consistent local/dev deployment.
Challenges I faced
My biggest challenge was latency and reliability in end-to-end processing. Early runs on longer videos were too slow, and I also hit UI data-sync issues like missing thumbnails, blank clip cards, and stale project state after deletion. I improved this by tightening the processing flow, fixing asset-path/data binding bugs, improving render state updates, and refining compare/export UX so results are clear and usable under hackathon constraints.
What I learned
I learned that AI quality alone is not enough; product experience depends on pipeline orchestration, asset consistency, and responsive UX. I also learned how important structured outputs are when integrating Gemini into real creative workflows. Most importantly, I learned that the best AI tools don’t replace creative direction, they accelerate it.
Log in or sign up for Devpost to join the conversation.