The AI Chapel is a spatial installation that proposes a new paradigm for human–AI interaction in the workplace. Rather than optimising for productivity, it optimises for reflection. The user enters a dedicated physical space, sits before an altar, and interacts with AI at a bounded time — the way you would not pray on a street corner, but go somewhere that holds that conversation differently.
The installation consists of three interconnected components: the physical chapel structure, the Conscious HUD displayed on a CRT screen below the altar, and the Confession HUD — a visitor-facing interactive activation powered by voice input and the Claude API.Conscious HUD
The Conscious HUD is built in TouchDesigner and displayed on the CRT screen. It operates through voice analysis and biometric input captured via microphone, generating real-time feedback across three dimensions:
Cognitive engagement(Teacher) — task complexity classification, depth-of-thinking signal** Emotional state**(Therapist) — vocal pattern analysis, dependency and attachment signals** Cognitive load**(Neuroscientist) — session duration, frequency of use, rest signal
The HUD displays a gradient-only visual with centred italic typography. No dashboard elements, no icons — only the signal. Three guard-case scenarios trigger interventions: Too Easy Tasks, Too Emotional, and Session Ending Score, each requiring a structured reflection before the session closes.Confession HUD
The Confession HUD is a visitor-facing activation layer built on top of the Conscious HUD. When a visitor approaches the altar, the webcam detects presence and the system initiates the interaction via the Claude API.
Interaction flow:
- Presence detected via webcam
- Gender neutral AI voice (ElevenLabs) welcomes the visitor
- Visitor speaks their AI confession aloud — microphone captures input, live italic transcript displayed on screen
- Visitor says "I am done" — system stops listening
- Claude API assesses the confession across the three saint perspectives (Teacher, Therapist, Neuroscientist) in a single continuous response, maximum 90 words
- Gender neutral voice delivers the response
- Session closes with: "You may leave. The chapel will remember."
The installation includes a 20-minute video narrative structured in five chapters:
Chapter I — Cloud Oracle(0:30): Oracular, inhuman AI voice introduces the dilemma** Chapter II — The Altar**(3:00): Gender neutral guide walks through the space, introduces the three saints** Chapter III**(voiceless — music only)** Chapter IV — The Prayer**(5:00): Call-and-response prayer, three movements per saint, tribal experience with two eyes-closed moments and a music peak at 3:05Chapter V(voiceless — ritual music)
Audio produced in ElevenLabs using SSML tags for precise silence timing. Post-processed in Adobe Premier with long-tail reverb and low-frequency hum for the oracular voice.
[Read more](javascript:void(0))