A localisation prompt builder turns an LLM into a transcreation partner rather than a literal translator. Word-for-word translation produces copy that is technically correct and subtly wrong — idioms that do not land, examples nobody recognises, the wrong currency, a tone that reads as rude or stiff. This tool builds a prompt that tells the model to adapt your content for a specific market while preserving what you actually meant.
How it works
You describe the content type (marketing copy, UI strings, support article, etc.), the target market, the audience segment, and the format. The builder assembles a prompt instructing the model to localise rather than translate: keep your intent and key messages, but adapt idioms and cultural references, adjust tone and register for the audience, convert dates, numbers, currency, and units to local conventions, and respect any length constraints the format demands. Crucially, it tells the model to flag decisions a human should make — like a price that needs re-setting rather than converting. The prompt is built in your browser; your content stays on your machine.
Tips and examples
Always name the market, not just the language — “French for Canada” and “French for France” diverge in vocabulary, formality, and legal phrasing. For UI strings, set the format so the model respects length limits; a button label that doubles in length will break your layout, and the prompt can ask the model to offer a shorter alternative. Treat flagged items as the point of human review: prices, legal claims, and culturally sensitive references usually need a person, and a good localisation prompt surfaces them rather than guessing. For ongoing work, lock in a glossary of brand terms that must not change and include it in the prompt so product names and taglines stay consistent across every market.