Pillar 02 · Research Program

AI Opera Docent
& Live Avatar

A conversational AI guide — with a live avatar — that introduces opera's works, performers, and history to audiences on their own terms, using natural voice and generative video.

Overview

A Guide for Every Listener

The AI opera docent is a conversational AI guide — deployable as a voice chatbot and, in its more advanced form, as a live avatar with synchronized video — that introduces users to opera's works, composers, performers, and history through natural dialogue.

The design philosophy is accessibility-first: the docent meets users at their level of familiarity, from complete novices to experienced opera-goers, adapting its explanations, vocabulary, and depth accordingly. Voice delivery using high-fidelity AI synthesis creates an experience that feels personal and human rather than transactional.

The avatar form — a real-time AI-generated video persona that speaks, reacts, and gestures naturally — extends the docent into an entirely new modality. Research examines whether embodied avatar interaction produces measurably different engagement outcomes compared to voice-only or text-based equivalents.

The system is deployed in partnership with Opera Verace Foundation and evaluated at collaborating opera institutions as an audience development tool offered to ticket purchasers before, during, and after performances.

Docent Capabilities

  • Real-time conversational response to open-ended opera queries in natural language
  • High-fidelity AI voice synthesis for warm, characterful vocal delivery
  • Live avatar video: synchronized face, speech, and gesture via generative video AI
  • Adaptive explanation depth calibrated to user familiarity
  • Knowledge base spanning opera history, works, composers, and performers
  • Integration with STAGE platform data on current productions and artists
  • Multi-language support for international and non-English-speaking audiences
  • Longitudinal conversation memory for returning users
  • Web, mobile, and in-venue kiosk deployment modes

Research Questions

RQ1 — Engagement

Does the docent increase attendance intent?

Pre/post measurement of ticket purchase intent among users who interact with the docent versus a control group.

RQ2 — Avatar Effect

Does a live avatar improve outcomes over voice alone?

Controlled comparison of voice-only vs. avatar modalities on engagement depth, return rate, and user-reported experience.

RQ3 — Retention

Do users return, and does repeat interaction compound impact?

Longitudinal study of repeat interaction rates and correlation with downstream attendance behavior tracked through STAGE.

Technology

System Architecture

🧠
Language Model
Large language model powering dialogue, knowledge retrieval, and adaptive response generation
🎙️
Voice Synthesis
High-fidelity AI voice via ElevenLabs, with custom voice persona calibrated for warmth and authority
🎭
Live Avatar
Real-time generative video avatar with synchronized facial animation, gesture, and expression
📊
STAGE Integration
Live connection to the STAGE data platform for current artist, production, and venue information