What it does

JeelizWeboji runs real-time face detection and expression tracking entirely in the browser via WebGL. It captures 11 facial expressions and head rotation on three axes from a webcam feed, then pipes that data into animated avatars built with Three.js or SVG. The neural network and all processing stay client-side.

The interesting bit

The README opens with a genuinely funny legal notice: Apple allegedly threatened IP action over a 3D fox demo in 2019, so the project now features a “cute raccoon” instead. The library also ships with a mesh converter that assembles morph-ready 3D models from separate .OBJ files—useful if you want to build your own expression-driven characters.

Key highlights

• Detects 11 facial expressions plus 3-axis head rotation
• Includes helpers for Three.js and SVG animation pipelines
• Works with bundled module imports or plain script tags
• Apache Cordova demo for native iOS (with a websocket hack for older versions)
• Third-party projects include hands-free mouse control, drowsiness detection, and a face-controlled piano

Caveats

• Requires HTTPS even for local development; HTTP blocks camera access in most browsers
• GPU performance directly impacts smoothness; underpowered graphics will stutter
• Beards, glasses, and bad backlighting degrade detection accuracy
• No official React/Vue/Angular demos included, though community boilerplates exist

Verdict

Worth a look if you’re building web-based avatars, accessibility tools, or novelty camera apps. Skip it if you need server-side processing or guaranteed accuracy across all lighting and hardware conditions.