How it is priced
Translation is priced per source word, per language. Review, which is a separate line and often omitted from a cheap quote, is priced per word or per hour. Machine translation with human post-editing sits below full human translation, sometimes substantially. Certified or specialist translation, which museums occasionally need for legal or repatriation-related text, sits well above it.
The variables that move a per-word rate are the language pair, the subject matter, and the turnaround. A common European pair is cheaper than a rarer one. Historical and technical content costs more than visitor information because it needs a translator who can handle the terminology. And a rush job costs more everywhere.
Per source word, per language, is the base unit
Review is a separate cost and is the line that gets dropped from cheap quotes
Machine translation with human editing costs less than translating from scratch
Rarer language pairs, specialist subject matter, and rush turnarounds all cost more
Estimating your own word count first
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. A ninety second audio stop is around 220 to 250. So a twelve stop guide is roughly 3,000 words, and a gallery of forty labels with a dozen panels is somewhere near 5,000.
Multiply by the number of languages and you have the shape of the quote before you receive one. It also shows you immediately where to cut: reducing from twenty stops to twelve saves the same amount in every language at once, which is a much larger lever than negotiating the rate.
Why printed translation costs more than the translation
Translating a panel is cheap. Producing the panel is not. Once the translated text exists it needs to be laid out, proofed in a language the designer probably does not read, printed, and mounted, and if it is wrong the whole cycle repeats. That production cost is usually several times the translation cost and it recurs every time the text changes.
It also caps you. A panel that has to carry three languages needs a bigger panel or smaller type, and at some point the exhibition design refuses. That ceiling is why museums with genuinely international audiences end up moving translated depth off the wall regardless of budget.
Translated text delivered digitally removes the production line entirely. The cost is the translation and the review, once, and correcting an error later costs the correction rather than a reprint.
The cost that recurs
Translation looks like a one-off and behaves like one for a permanent collection, as long as nothing changes. What generates recurring cost is change: a corrected date, a rewritten panel, a new acquisition, a temporary exhibition. Each of those is a small translation job in every language you publish, plus the review, plus the production if it is printed.
Museums consistently underestimate this because the first translation is a project with a budget and the subsequent ones are not. Deciding up front how many languages you can afford to maintain, rather than how many you can afford to launch, produces a more honest number and usually a shorter list.
Frequently asked questions
How much does museum translation cost?
It is priced per source word, per language, with review as a separate line. The number you can control is the word count: a twelve stop audio guide is around 3,000 words and a gallery of forty labels plus panels is nearer 5,000, and each language is a multiple of that.
Is machine translation cheaper?
Yes, meaningfully, when it is used as a first draft that a human editor corrects. Publishing raw machine output is not a cost saving so much as a risk transfer to your visitors, and for culturally sensitive content it is not an option.
Why is translating printed panels more expensive?
Because the translation is the small part. Layout, proofing in a language the designer cannot read, printing, and mounting cost several times the translation, and the whole cycle repeats every time a word changes.
Does adding a language cost the same each time?
Roughly, for translation and review. Everything else is already paid for, so the marginal cost of a fourth language is much lower than the cost of the first, provided the delivery does not involve reprinting anything.
How do we reduce translation cost without cutting languages?
Cut the word count and lock the source text before translation starts. Fewer, better stops save money in every language at once, and every edit made after translation begins is paid for again in each one.