QuickReels | Devpost

Demo - Main Page
Demo - Video Submission
Demo - Processing
Demo - Ready
Title
Description
Overview
Current Situation
Problems
What we do
Key Features
How does it work
Our AI
Use cases
Creators
Educators
Sports
Possibilities

Inspiration

Our inspiration comes from the question: How does a creator turn an existing video into a reel format? We want to solve that. No more wasting time deciding what to highlight. No more cropping. No more resizing. No more editing. QuickReels does it all for you with one click!

What it does

Let’s you pick the best moments from your video

Automatically resizes any video to 9:16 aspect ratio
Analyzes the content on screen and detects camera and character movement
Highlights the most relevant action (for example, the person talking) in the centre
Handles camera panning, transitions, and multiple-people tracking automatically

How we built it

We built our algorithm using Google's MoveNet model in TensorFlow to calculate people's coordinates, along with our own smoothing curve math to normalize them. For the web version, we used Vite and React. This program works in both CLI and web versions.

Challenges we ran into

We tried many different algorithms, from large rulesets to real physics simulations, to find the best camera handling simulation. Here is a demo of our simulations: YouTube Link