CALIBRATING LAB
Loading interface
Squid Game AI
Sep 2025

Squid Game AI

Red Light, Green Light - Detected by Machine

OpenCVMediaPipePythonPose EstimationReal-Time

“The game's rule is simple: don't move. Enforcing it with a machine is the challenge.”

01

The Game

Inspired by the Red Light, Green Light challenge from Squid Game, this project uses real-time body pose detection to monitor player movement via a webcam. During the Green Light phase, players can move freely. When the system switches to Red Light, any detectable movement in the player's pose results in elimination, just like the show.

02

Pose-Based Movement Detection

MediaPipe's pose estimation model tracks 33 body landmarks per frame, including joints from the hips, shoulders, elbows, knees, and feet. Movement is detected by computing the frame-to-frame delta of a weighted set of key landmarks. If the cumulative displacement exceeds a calibrated threshold during the Red Light phase, the player is flagged as having moved. The threshold was tuned to ignore minor camera noise while catching clear body movements.

03

System Design

The game loop manages phase transitions (green-to-red with randomised timing), visual feedback overlaid on the camera feed (countdown timers, colour-coded borders), and a real-time score tracking eliminated vs. surviving players. A sound system cues the phase changes, matching the tension of the original game. The result is a surprisingly effective recreation that demonstrates how pose estimation can be used as a game input mechanism with essentially zero additional hardware.

Project Gallery

3 items
Squid Game AI media 1
Squid Game AI media 3
Squid Game AI media 2

Explore More Projects

Swipe or click arrows to see other work

Built and designed by Tharsshen Govikrishnan · © 2026