Real-time animation
Sub-second render pipeline drives lip-sync, micro-expressions, and gaze. The persona listens and reacts while it speaks.
- 60fps photoreal or stylized
- Gaze tracking via optional camera
- Idle gestures, not a frozen headshot
Echo Persona brings historical figures, fictional characters, and original hosts to life as animated AI companions, rendered in real time on any screen in your venue.
Every persona begins with a character bible, co-authored with your team — curators, brand leads, or creative directors. Primary sources, permitted anachronisms, speech patterns, and refusals all live in one document — and the avatar never leaves it.
Conversations run in real time across multiple languages. Visitors lean in because the character leans back.
Bounded knowledge base. In-voice refusals. Sign-off by your lead on every public claim.
Real-time lip-sync, gaze, and gesture. No prerecorded video, no uncanny loops.
Wall panel, kiosk, hologram, iPad. The persona follows the visitor.
Personas arrive with a worldview, a voice, and a body. Each component tuned by specialists, then tuned again by the institution.
Sub-second render pipeline drives lip-sync, micro-expressions, and gaze. The persona listens and reacts while it speaks.
A bounded corpus per persona (archival documents, curator notes, exhibition text, character lore) with explicit refusals for everything outside the frame.
Voice in, voice out. Personas code-switch on cue, and historical figures speak in their mother tongue when the visitor does.
Each exchange lives in the moment, but visitors can leave with a record of it. A QR scan at the end of the session captures the conversation as a takeaway they own.
Off-topic probes, adversarial prompts, and historical atrocities are handled with policies written by your curators.
Every conversation is a visitor research study. Dwell, topic clusters, sentiment, and drop-off surface in a curator-readable dashboard.
Pointe-à-Callière asked us to revive Joe Beef, the 19th-century tavern keeper whose knack for finding meat in any season earned him his nickname. Modelled in 3D from John Henry Walker's 1875 portrait, grounded in verified archives, speaking French or English.
See the case study
"The beauty of this new character is that he goes beyond simply reciting content. He transmits emotion, and shows humor, engagement, and sensibility in what he says."
Émilie Prévost, Project Manager, Pointe-à-Callière
3h
Of conversation every day
90s
Average per discussion
10k+
Questions answered each month
A team from Dpt. pairs with your lead — curator, brand director, or creative lead.
We read the archive, brand guidelines, or character bible with your lead. What the persona knows, and what it refuses to guess.
Week 1
Voice cast from period recordings or a character brief. Likeness create from period portraiture or IP assets.
Week 2
Red-team the persona against twelve hundred visitor prompts, then sit with your lead to tune refusals and register.
Week 3
Install on the floor. Week one readout. Monthly drops. Your persona gets sharper with every conversation.
Week 4 →
Can't find it? Write to bonjour@dpt.co and a human replies.
Any figure with a sufficient historical or lore record. Custom personas are built from your archives, brand guidelines, or character bible and vetted before launch. The persona never speaks outside its sources.
Yes. Personas are built from a character bible (voice, mannerisms, worldview, permitted knowledge). Theme parks, brands, and entertainment IP use original personas for hosting, narration, and signature storytelling, alongside historical figures reconstructed from archival sources.
Any display with a microphone and a network connection. The render runs in the cloud; on-venue hardware is a thin client. Typical installs are a vertical 55″ to 75″ panel with a directional mic and an optional camera for gaze tracking.
Each persona is bounded by a curated knowledge base and a refusal policy written with your team (curator, brand lead, or creative director). When asked outside its scope, the character declines in-voice — "I couldn't say" — rather than inventing. Every public deployment is signed off by your team.
Conversations are ephemeral by default; a visitor can opt in to a continued session via QR. Data is GDPR- and CPPA-compliant, visitor-held.
A subscription per persona per venue, plus a one-time build for custom characters. Pricing is conversational : most engagements land between a small activation and a large permanent installation. Write to us for a quote.
Thirty-minute call with an Echo lead.