Enhanced

Your content,
out loud.

The agent speaks its answer as it writes it. Words light up as they are read, so your visitor reads and listens at once. In 80 languages, in a voice you built.

Hear it on your content
Enhanced interface reading its answer aloud, on The Mountain Climber by Fernand Leduc at the MNBAQ
80 languages its own character reads aloud

Read and listen

The answer speaks itself.

Text and voice, together

Each word brightens the moment it is spoken. Follow with your ears, your eyes, or both.

Key facts stand out

The agent emphasises what matters inside its own answer, so a skimming visitor still gets it.

Silence is one tap away

A visitor in a quiet room mutes the voice and keeps reading. Nothing else changes.

  • Key facts emphasised inside the answer
  • Images and video when they help
  • One tap to mute the voice and just read
  • Keyboard or microphone, the visitor decides

Voice and character

Not a default voice. Yours.

Most agents sound like the model that runs them. Yours should sound like the institution that speaks through it.

We build a character before we build an interface: who it is, what register it uses, what it declines to answer, how it addresses a child and how it addresses a curator. That method comes from nine years spent inside cultural institutions, not from a settings panel.

See the engine underneath

The character brief

Who it isa figure from your own historya work that speaks for itselfa guide you inventan ambassador already tied to youno character at all, by design
How it speakswarm and plainprecise, almost formalplayful and quickpatient, one idea at a time
What it calls youformal, alwaysinformal with children, formal with adultsby first name, if that is your house style
Its voicechosen from a librarycloned, with written consentone voice per language, same character
What it refusesAnything your sources do not support
Down to the phoneticsProper nouns forced syllable by syllable, so the voice never stumbles

Every line above is a dial, not a preset. We set them with you, your teams sign them off, and no two agents come out alike.

One question, three visitors

The same question.
Three different answers.

Before the first message, the visitor says how they want to be spoken to. A specialist, a family and a visitor who needs clarity never get the same reply, and none of them has to ask for it.

Visitor asks The Captain: “Were you ever afraid out there?”

Default

Richer vocabulary, deeper insight

Fear was ballast, not cargo. I learned to read a sky the way a clerk reads a ledger, and the years taught me the sea punishes hurry long before it punishes fear.

Family

Simple words, playful

Terrified! One wave took my hat clean off my head. Do you want to know whether I ever got it back?

Accessibility

Clear and concise

Yes, often. The sea is dangerous. I learned to watch the sky, and to be patient.

The visitor chooses how the agent communicates

The visitor makes that choice once, on opening, in one tap. It carries through the whole conversation and into every language the agent speaks.

Field results

A visitor who listens keeps their eyes on the artwork.

Reading on a screen pulls attention out of the room. Voice gives it back.

0average time in front of an equipped artwork, against five seconds beforeMNBAQ, 50 artworks equipped
0staff hours freed every year on visitor questionsChâteau de Chillon, three languages
0more engagement than a conventional mobile appChâteau de Versailles, 20 statues and fountains

Your identity

It is your agent, in your colours.

Your typography, your palette, your avatar, your welcome screen, your name on it. Visitors meet your institution, on a page that belongs to you.

Saint Savina, slate blue interface
The Metropolitan Museum of New YorkSaint Savina
Kiran Nadar Museum of Art, terracotta interface
The Kiran Nadar Museum of New DelhiMadhya Pradesh Tiger
The Alimentarium Museum of Vevey, green interface for La Fourchette
The Alimentarium Museum of Vevey« La Fourchette »

The foundation

The same engine and the same console behind all three levels.

Essential, Enhanced and Immersive change what the visitor sees. What answers, and what you steer it with, never changes. Follow one question through it.

Your sourcesYou load
The knowledge baseYou edit
The engineIt answers
The evaluatorIt checks
The visitorIt speaks

And it comes back to you. Every conversation turns into structured data in the console: what audiences asked, what they failed to find, when, and in which language.

Your sources

Whatever you already have, in the format it already exists. Nothing to rewrite for the machine.

  • PDF and Word
  • Web pages
  • Spreadsheets
  • Images
  • Your collection database

The knowledge base

The console is yours. Change an answer, publish it, and see the result immediately, with no developer in the loop.

  • Edit and publish yourself
  • Preview and test instantly
  • One base, all three levels
  • No technical skill required

The engine

Several models under one orchestration layer, tied to no single provider. It answers from your sources and nothing else.

  • Only what you approved
  • No provider lock-in
  • Visual recognition on request
  • A work identified from a photograph

The evaluator

Every answer is scored before it reaches anyone, in each of the languages you serve.

  • Faithfulness to your sources
  • Tone
  • Safety
  • Scored in every language

The visitor

They ask in their own words and get your answer, read aloud as it is written, in the character you approved.

  • Text and voice together
  • 80 languages
  • Same character in each
  • Nothing to download

Next step

Hear it on your own content.

Send three documents and one artwork. We come back with a working agent that speaks, in the character you approved, in the languages you need.

Book a demo