Shadowbrook - A Voice You Can Look At
A face, a voice and a memory, in a room. Shadowbrook puts a realtime voice model inside a three-dimensional person you can look at, so a conversation feels like talking to someone who is there.
The Challenge
People will say things to someone who is present that they will not type into a box, and they will stay with the conversation. A voice model on its own is a waveform. The question this project asks is whether a face, a room and a memory change what people are willing to say, and how long they stay.
Our Solution
We built an invite-only web app. OpenAI Realtime or Gemini Live speaks; a Microsoft Rocketbox character lip-syncs in the browser and gestures from a library of recorded clips, standing or sitting in a photographed room. The character can recall what it has been told, and an invite can arrive with those facts already written in: the family, the dog, the routine. A locked invite opens one conversation the guest cannot reconfigure. A researcher invite opens the studio.
Key Features
Results & Impact
Project Details
Technologies Used
Planning something similar?
We work as an embedded product and engineering partner, from first commit to launch. Tell us what you're building.