Speak to every visitor in their first language.

More than forty languages are supported. Each institution activates up to ten at a time and can swap the set whenever it likes — most start with some mix of English, Spanish, French, German, Italian, Portuguese, Chinese, Japanese, Korean, and Arabic, then adjust to who actually walks through the door. The ten-at-a-time cap isn’t a capability limit; it exists because no institution can realistically vet forty languages. If you genuinely need more than ten active, tell us.

You decide. Convo produces the translations, but a translation only goes live when a person at the institution marks the script as approved. Most museums route translation review to whoever already vets multilingual wall text — an education manager, a docent, or an outside consultant for the languages no one on staff reads. We don’t pretend the platform replaces that judgment.

The difference is structural. Studio production charges per language per minute of finished audio — adding Spanish doubles the bill, adding ten multiplies it. Convo doesn’t charge per language: your active languages come bundled, and adding, swapping, or updating them after launch doesn’t trigger a new production cycle. The cost lives in the platform, not the language count.

Yes. The language picker lives in the visitor web app and follows them across stops. A family with a Korean-speaking grandparent and an English-speaking grandchild can hand the phone back and forth and the tour state persists. Questions asked of the visitor guide are answered in whatever language the question was asked in.

Most of the time, yes — but this is exactly the kind of thing review exists for. Named entities, transliterated names, the convention your institution uses for a particular artist or dynasty: these are decisions, not errors. When the translation surfaces something the institution has a house style for, you edit it on the script and the change applies on the next regeneration. We don’t pretend to know your house style; we make it cheap to enforce it.

Two paths. Either contract a reviewer for that language — the work is small enough (a few hours, total, for a typical pilot) that it’s within reach of most education budgets — or limit yourself, at first, to the languages someone on staff can vet. Better to ship five carefully reviewed languages than ten unreviewed ones. The platform doesn’t force the question.