Computer VisionBrowser-basedMediaPipe Face LandmarkerFace meshAR attachment pointsStudent lab

Face Landmark and Expression Studio

Detect a full MediaPipe face mesh from a webcam or uploaded image, highlight key facial regions, and study how AR filters attach to tracked facial points.

This page extends the Face Filters Studio into a more educational lab. Instead of only rendering effects, it shows the facial landmarks that make those effects possible: dense mesh points, eye regions for glasses or eyeliner, mouth landmarks for lip effects, nose landmarks for stickers, and the face outline for masks or beauty filters.

What MediaPipe predicts

MediaPipe Face Landmarker predicts a dense face mesh, not just a few feature points. That means students can inspect the overall facial surface plus focused regions such as eyes, lips, nose, and face oval.

How AR filters attach

AR filters use stable anchor points such as eye corners, the nose bridge, mouth center, and face outline. Once those points are tracked, overlays like glasses, crowns, blush, or masks can stay attached as the face moves.

Why this page is educational

This studio turns the same landmark idea behind the Face Filters Studio into a step-by-step teaching page: mesh points, highlighted facial regions, blendshape-based expression cues, and simple geometric measurements.

Upload image

Model: Loading MediaPipe Face Landmarker...
Mode: Live webcam tracking
Status: Start the camera to see how the face mesh drives AR-style attachments.

Loading MediaPipe Face Landmarker...

AR Attachment Map

Eyes and irises

Useful for glasses, eyeliner, lashes, gaze cues, and sparkle effects that follow eye movement.

Mouth and lips

Useful for lip tint, smile effects, speaking avatars, and mouth-open interactions.

Nose landmarks

Useful for stickers, animal noses, blush placement, and central face alignment.

Face oval

Useful for masks, contour overlays, beauty effects, and full-face tracking boundaries.

Mesh Summary

Faces

Mesh points

Blue points mark the eye regions, pink points mark the mouth, orange points mark the nose, and green marks trace the rest of the face mesh.

Basic Expression Cues

These are lightweight cues derived from MediaPipe face blendshapes. They are useful for teaching, but they are not the same as a full emotion classifier.

Expression-style cues will appear after the first face is detected.

Top Blendshapes

Blendshape scores appear when MediaPipe returns a tracked face.

Geometric Measurements

Eye distance— px

Mouth width— px

Mouth openness— px

Nose length— px