The problem is space, not language
A panel that carries 120 words in English cannot carry 120 words in four languages and still be readable from a metre away. Most museums hit this at design stage and quietly cut the language list back to one, which is how a collection ends up in a single language by accident rather than by decision.
Separating the two jobs is what breaks the deadlock. The printed label stays short and factual: what it is, who made it, when, and one line on why it is here. The full interpretation moves somewhere with no space limit, reached from a code on the label. After that, adding a language stops being an exhibition design problem, because nothing physical changes when you add one.
The stages of a translation project
Stage 1
Decide what gets translated
Not everything, and not in the order you would guess. Orientation comes first: how to get around, what is included, what the rules are. A visitor who cannot read the wayfinding is having a worse visit than one who cannot read an object label.
Then the objects people already stop at. Staff know which ones those are because they answer the same questions about them every day, and twelve of those done properly beats the whole collection done thinly.
Stage 2
Choose the languages honestly
National tourism figures describe the country, not your venue, and nationality is not the same as language need. Your front desk knows which visitors arrive with a translating companion and which ones photograph panels to read later, which is better evidence and free.
Three or four languages is a realistic first list. Start with the local language, which is where most listening happens, then add the ones you have evidence for.
Stage 3
Lock the source text
Every edit made after translation begins is paid for again in every language, and twice over if that language has already been voiced. A named person signs the source off and later changes go on a list for the next revision.
This is the single largest cost control in the project and the one most often skipped, because assigning that authority is uncomfortable.
Stage 4
Translate, then have someone check it
You need two people per language: one to translate and one to review. Publishing text nobody on staff can evaluate means the first person to evaluate it is a visitor who speaks it.
Reviewers do not have to be professional linguists. Community organisations, language schools, and inbound operators catch the errors that matter, which are names, tone, and accidental rudeness rather than fine grammar.
Stage 5
Publish it where space is not the constraint
A panel carries one language comfortably and two at a push. Put the short factual label on the wall and the full translated interpretation behind a code on it, and the number of languages stops competing with the exhibition design.
Publish text and audio together. The same translation serves visitors reading it, visitors listening while they walk, and visitors who cannot read the panel at all.
Stage 6
Keep them in step afterwards
Six months later somebody corrects a date in English and the translations quietly disagree with it. Nobody notices, because nobody on staff reads them.
Keeping every language of a stop in one place is what makes that visible. Translations living in separate documents from an agency drift language by language until the guide is wrong in ways you cannot see.
How visitors will reach the translation
There are three practical routes, and the choice decides how many languages you can afford and what a correction costs later.
Printed panels and labels
Works well for: Visible to everyone in the room with no device, no scanning, and no signal.
Costs you: Space. A second language roughly doubles the text and a third makes the panel unreadable, and every correction means layout, proofing in a language your designer cannot read, printing, and mounting again.
Printed booklets or laminated sheets
Works well for: Cheap to produce, easy to hand out, and works where a phone will not.
Costs you: They go missing, wear out, and get out of date silently. Staff have to hand them over, which means every translated visit depends on someone noticing the visitor needs one.
Behind a QR code on the visitor's phone
Works well for: One code carries every language you publish, in text and audio, and a correction costs the correction rather than a reprint.
Costs you: Needs usable connectivity where visitors scan, and a printed sign that makes it obvious the translation exists at all.
Journey Pal takes the third route. One code beside the object opens the stop in the language the visitor's phone is set to, as text and audio together, with a switcher for anyone who wants a different one. Adding a language is a content change, so the signage never has to be touched again.
Where machine translation fits
Machine translation is now good enough for a large amount of museum content and still unsafe for a specific, predictable subset. Plain, factual, present-tense text translates cleanly: hours, directions, accessibility information, the rules about photography. A museum that machine-translates its visitor information is using a free tool sensibly.
The failures cluster in exactly the content a museum cares most about. Proper nouns get translated when they should have been left alone. Historical terminology gets modernised, which can turn a careful description into an offensive one. Register collapses, so text written with dignity comes out casual.
The dangerous property is confidence. Machine output never hesitates and never flags the line it got wrong, so a museum with nobody on staff who reads that language cannot tell which sentences are safe. The error gets found by a visitor, in the gallery, and it is the visitor you were most trying to welcome.
- Safe: hours, directions, ticketing, accessibility information, rules
- First draft only: interpretive text, then edited by a speaker of the language
- Not without a qualified reviewer: memorial text, contested history, personal testimony
- Not at all: Indigenous and minority language content, where quality is weakest and stakes are highest
What translation actually costs
Translation is priced per source word, per language. Review is a separate line and is the one that disappears from a cheap quote. Machine translation with human post-editing sits below full human translation; specialist or certified translation sits well above it. Rarer language pairs, historical subject matter, and rush turnarounds all push the rate up.
You can size the job before asking anyone. A short object label is 50 to 80 words, a full interpretive panel is 150 to 250, and a ninety second audio stop is around 220 to 250. Multiply by your language count and you have the shape of the quote in advance.
The larger lever is not the rate, it is the word count and the delivery. Cutting from twenty stops to twelve saves in every language at once. And translating for a printed panel costs several times the translation itself, because layout, proofing, printing, and mounting all follow it, and the whole cycle repeats whenever a word changes.
Translation as accessibility, not just tourism
Translation is usually budgeted as a service to international visitors, which undersells it. The same work reaches residents who speak another language at home, visitors who read slowly or find dense panels tiring, and anyone who would rather listen than read while walking.
Publishing every stop as text and audio together is what makes that hold. A visitor who is deaf reads the transcript, a visitor with low vision listens, a visitor reading in a second language can do both at once and understand more than either alone. None of that costs extra once the translation exists, and it is invisible if the translation only exists as printed text on a wall.
It is also worth saying plainly on the signage. A visitor who does not read English will not know the guide exists in their language unless the sign tells them in their language, and that single line is one of the cheapest things in the whole project.
The mistakes that cost the most
- Starting translation before the source text is signed off, which means paying per word twice in every language.
- Committing to a language with nobody available to review it.
- Putting every language on the panel, then discovering the type is too small to read.
- Machine-translating culturally sensitive or Indigenous language content without a qualified reviewer.
- Letting translations live in separate agency documents, so a correction to the English never reaches them.
- Translating the object labels but not the wayfinding, which is what a visitor needs first.
- Publishing the translation but not saying so on the signage, in that language.
Go deeper on any part of it
Museum translation software without crowded walls or app downloads
Museum translation software for multilingual visitors, QR code delivery, audio interpretation, and easy content updates without crowded signage.
How to translate exhibit labels
How museums translate exhibit labels and wall panels without doubling the text on the wall, which parts to translate first, and why panel space is the real constraint.
Is Google Translate good enough for a museum?
An honest look at using Google Translate for museum labels and guides: where it genuinely works, where it fails in public, and what to do about the gap between them.
What museum translation costs
How museum translation is priced, what makes a quote move, why printed panels cost far more to translate than digital text, and where the recurring cost hides.
How to translate an audio guide
How to take an audio guide from one language into several: locking the source script, what changes when text becomes speech, and reviewing a language nobody on staff reads.
Which languages should you translate into?
How to choose which languages to translate a museum guide into, why visitor nationality data misleads, and why the local language is usually the biggest win.
Frequently asked questions
How do museums translate exhibits?
Usually by keeping the printed label short and factual and putting the translated interpretation behind a code on it, because panel space runs out after one or two languages. The translation is commissioned per word, reviewed by a second speaker, then published as text and audio together.
How many languages should we translate into?
Three or four to start: the local language plus the two or three you have real evidence for. Adding more later is a content change rather than a rebuild, so there is no advantage to over-committing at launch.
Can we use Google Translate for museum text?
For operational text such as hours, directions, and rules, yes. For interpretation, use it as a first draft that a speaker of the language edits before it is published, and keep it away from culturally sensitive and Indigenous language content entirely.
How much does museum translation cost?
It is priced per source word, per language, with review as a separate line. The number you control is word count: a twelve stop audio guide is around 3,000 words, and a gallery of forty labels plus panels is nearer 5,000.
Do we have to reprint signage to add a language?
Not if the translated text lives behind a code rather than on the panel. One QR code serves every language published behind it, so the printed sign stays exactly as it is however many you add.
What about visitors who do not speak English?
They are the visitors translation helps most, and the ones a printed English panel excludes completely. A translated guide lets them explore independently rather than relying on a companion to translate wall labels for them.
Should the audio be translated as well as the text?
Publishing both is the point. Text serves visitors who prefer to read or cannot hear, audio serves visitors who are walking, looking, or find reading in a second language tiring, and the same translation pays for both.
Who should review a translation?
Anyone fluent who did not produce it. The errors that matter in museum text are names, tone, and something that reads as unintentionally rude, and a community organisation or language school will catch those.
Want the translations delivered without the signage problem?
Journey Pal publishes your translated text and audio behind a single QR code, opens each visitor in the language their phone is set to, and lets you add a language or fix a line without reprinting anything. You decide which languages are worth publishing, which is the part that needs your knowledge of your visitors.