Inspiration

We've all been there: hunched over a laptop at 2am, eyes glazed, definitely not working. We wanted a companion that notices when you've checked out or when you're struggling, and actually does something about it. Not a notification. Not a pop-up. Something physical.

What it does

"Tough Love" runs a continuous webcam loop, processing each frame with MediaPipe's face mesh to extract 468 facial landmarks. From those landmarks, we compute three signals:

  • Mouth droop: compares the Y-position of mouth corners relative to the nose tip to detect frowning
  • Head tilt: measures the vertical offset between nose and forehead landmarks
  • Face centering: checks horizontal nose displacement to catch when the user looks away

If the face is absent for more than 3 seconds, we classify the user as distracted. Each state triggers a robot arm sequence via the Cyberwave API and a spoken response via ElevenLabs, running in parallel threads so the arm moves while the voice plays.

A 5-second cooldown per state prevents the arm from going haywire during transitions.

Built With

Share this project:

Updates