LearnAR
Lessons lived in the place they teach about — kitchen, machine shop, museum, gym.
What it does
Hold the camera up to a thing. Say what you want to learn about it. The instrument identifies the object, drafts a walkthrough, and anchors each step to the part of the object the step concerns. The avatar narrates as you go.
The lessons are composed for the thing in front of you, in the room you're in, in the language you've set.
How it works
The intent — what you want to know — goes to Gemini along with a frame from the camera. Gemini returns a structured walkthrough: ordered steps, each with an anchor target (which part of the visible object), a narration script, and an optional callout overlay. The instrument anchors each step in space and waits for you to step through.
What it's built on
ARKit world tracking. Google Gemini for intent → structured walkthrough. The HowToKit step-engine for state management. Azure Speech for the narration. The companion avatar for voice and presence.