How Convo works.
Two things, working together. A platform that drafts, voices, translates, and publishes audio tours from your reference materials. And tours your visitors can talk to — at any stop, in any language.
From your reference materials
to a published tour.
Drafted in ninety seconds. Edited line by line, or by describing the change. Voiced in ten languages. Published behind a QR code at the entrance. Updated as fast as your collection changes.
Bring everything you already have.
PDFs. Word docs. CSVs. Wall-card text. Exhibition essays. Catalogs. Curatorial notes. Convo reads them, indexes them, and uses them as the source of truth for every draft and every visitor answer.
Bulk-import a whole collection from a CSV — image, accession, label, description per row. Toggle scanning per item to let visitors point their camera at an object and land on its stop.
Drafted in ninety seconds. From what you uploaded.
Convo reads your sources and proposes a script for every stop — typically a hundred and fifty to two hundred words, ready to voice in under two minutes.
The first draft is a starting point, not a final. Regenerate as many times as you want; nothing reaches a visitor until you approve it.
Edit the way you write. Or describe the change.
Edit a script directly — like a Google Doc. Or open the assistant and describe the change in your own words. “Less academic.” “Lead with the back of the sculpture.” “Rewrite for visitors aged eight to twelve.”
Convo proposes a diff. You apply or revert. Versions are kept; you can restore an earlier draft any time.
Ten languages from one source.
English, Spanish, French, German, Italian, Portuguese, Chinese, Japanese, Korean, Arabic. Approve a stop in your source language; Convo regenerates the rest in about a minute and revoices each one.
Six voices configurable per institution — three female, three male — with tone and register set in org settings. Languages ship together when you re-publish, so they don’t drift.
A correction is no longer a project.
Re-edit a stop, click publish — every language re-voices and goes live in about a minute. A typo, a corrected attribution, a rotating exhibit added mid-cycle: same path, same speed.
The cost of an update used to mean a curator chose between accuracy and shipping. That trade-off is gone.
The draft for every stop in the wing is ready by lunch.
By the time you replied, the corrected stop was live in ten languages.
You wouldn’t have, before.
Once a tour is live, every visitor
follows their own curiosity.
Chat or voice. At any stop, in any language. Answers come from your reference materials — never the open internet. The same tour adapts to a child, an art historian, and a tourist on the same afternoon.
Two ways to ask. Same answer.
Type a question; an answer comes back with a tap-to-speak option. Or hold the mic, ask in your own voice, and hear the reply spoken in the same voice as the tour. Language is auto-detected per visitor.
Suggested prompts at every stop are curator-controlled, so visitors always have a reasonable place to start.
You decide what visitors are invited to ask.
Convo proposes prompts based on your sources. You pick, edit, and order them. Whatever you save shows up as the tap-to-ask chips at that stop.
A simple, visible form of scope control — and the easiest way to seed the conversations you want visitors to have.
Point the camera. Be at the stop.
Open camera in the visitor app, point at an object — Convo matches against your collection and routes to the right stop. No QR sticker next to every label.
Curators enable scanning per item; image vectors are computed at upload. Works alongside QR codes; visitors can use whichever they land on first.
What visitors do,
and what they ask.
Three things in one place — engagement, themes, and every individual question. Reviewable, exportable, the kind of insight that goes in a board memo.
Built for the people who have to defend it.
The visitor guide answers from your reference materials, not the open internet. When it can’t ground itself, it declines.
Every question and answer logged. Searchable, exportable, flaggable.
Your materials are yours. Not used to train models, not shared with other institutions.
A three-status workflow. Nothing visitors see is unreviewed.
Viewer, editor, admin. Invite stakeholders without giving up control.
Four plans, no asterisks.
Asked and answered.
Thirty minutes.
Your collection,
in your voice.
Bring one gallery and a handful of reference materials. We’ll show you what it would sound like — in English, and in two more languages of your choosing.