gazijarin/itsgiving

What it solves

It's giving allows users to trigger visual meme overlays on their webcam feed in real-time based on their facial expressions and hand gestures. It solves the problem of static video calls by adding interactive, gesture-driven reactions that can be broadcast to others via a virtual camera (e.g., in Zoom or Microsoft Teams).

How it works

The application uses MediaPipe to track 478 facial landmarks, 52 blendshapes, and hand/body positions. It calculates face-relative geometry to ensure the system works regardless of the user's distance from the camera.

To handle the fact that different people have different resting facial expressions, the project includes a calibration mode. This mode records the user's neutral face for seven seconds to establish a mean and standard deviation for each expression channel. It then calculates a "z-score" (sigma), measuring how far a current expression deviates from the user's specific neutral baseline, making the detection reliable across different face shapes.

Who it’s for

• Content creators and streamers who want interactive visual reactions.
• Remote workers or students looking to add humor and engagement to video meetings.
• Developers interested in computer vision and gesture recognition.

Highlights

• Virtual Camera Support: Integrates with OBS or v4l2loopback to appear as a standard webcam in meeting apps.
• Personalized Calibration: Uses statistical normalization (z-scores) to adapt to individual resting faces.
• Customizable Assets: Users can easily swap memes by replacing images or GIFs in the assets folder.
• Extensible Pose System: Provides a framework for developers to add new gestures by defining geometry-based rules in the decide() function.
• Scale-Invariant Tracking: Normalizes distances by face width so gestures are detected consistently at any distance.