Meething : Machine-Learning Camera

Machine-Learning powered Virtual MediaDevice extension for any browser-based conferencing service.

Behind the scenes, the actual camera frames are being processed by PoseNet and FaceMesh producing coordinates used to animate an SVG Character which replaces the camera output, never streaming your real image.

Usage

To test the model without installing any extension, run the following or use glitch:

npm install
npm run build
npm start

Usage as Extension (alpha)

Nothing to compile if you want to use the extension - Follow the instructions to install in your preferred browser:

Chromium

• download or clone the repo
• go to chrome://extensions
• enable Developer Mode
• Load unpacked
• Browse to the dist folder with the extension (where the manifest.json file is)

Firefox

• download or clone the repo
• go to about:debugging#/runtime/this-firefox
• Click "Load Temporary Add-on..."
• Browse to the dist folder with the extension and select the manifest.json file

Usage without Extension

If you want to use this without extension, try our decentralized conference project Meething

Credits

This humble hack-speriment would not be possible without the following projects:

• Virtual Browser Camera shaders by spite
• Pose Animator by yemount
• Meething by Team Meething