Magic Words VR

Voice-controlled make-believe environment in virtual reality


Magic Words is built entirely with web technologies, it uses Mozilla's A-Frame WebVR framework and can run on anything from phones and desktop browsers to Google's Daydream and Oculus Rift. It requires no additional attachments or peripherals and was designed for maximum compatibility and accesibility.

Realtime collaboration is easy, just share the unique URL of the project and any partners who enter appear in the shared scene. Multiple people can request models, manipulate them and create a world together.


Magic Words VR is a virtual reality world-conjuring environment controlled by the intuitive interfaces of speech and movement. It is an experiment in building authoring tools for VR that require little to no training and leverage familiar human interactions (such as moving around and talking to each other).

Magic Words integrates with the Google Poly 3d models collection and works by contextual voice commands: "Draw an elephant please" will query the Poly API for "elephant" models and place it in the point the user is looking at. Look at the elephant, and you'll be able to control it further: "another one", "bigger", "smaller" and "follow me" are just some of the supported interactions.

Bill of Materials


  • Google Daydream
  • Google Pixel 2


  • A-Frame
  • Three.js
  • Google Poly API
  • Web Speech Recognition API
  • WebVR