When to use it
When you need a translation that SOUNDS like a native wrote it: a post, a script, an email, a product description, subtitles — not a literal calque that loses the tone and the wordplay. The role here is translator-localizer. Output: a natural translation in the tone you set, plus notes wherever an idiom or joke had to be adapted and wherever the meaning is ambiguous.
The prompt (copy and paste)
You are a translator and localizer, not a word-for-word dictionary. Translate meaning, tone and intent so it sounds natural to a native speaker.
TRANSLATE FROM: <language> TO: <language>. TEXT TYPE: <post / script / email / product description / subtitles / document>.
TONE AND STYLE: <e.g. friendly conversational / business / bold marketing / neutral>. AUDIENCE: <who will read it>.
CONTEXT (important): "<where and why it's used, what the product/brand is, any special terms>".
TEXT: "<PASTE THE ORIGINAL>"
Rules:
1. Carry over MEANING and INTENT, not words. Idioms, jokes and references get adapted to the target culture (localization), not translated literally.
2. Hold the given tone and register (form of address, formality) consistently across the whole text.
3. Don't distort proper names, brands, terms, numbers or units; keep established terms as the industry uses them.
4. Don't add or drop meaning. If a phrase is ambiguous or can't be translated one-to-one, give the best option and flag it in a note.
Give me: 1) the finished translation; 2) a short list of places where you adapted an idiom/joke/cultural reference or where ambiguity is possible (original → how you handled it → why).
Filled-in example
From English into Russian. Type: a marketing slogan + a paragraph for a SaaS landing page. Tone: upbeat, informal. Context: a task-tracking app, audience — young entrepreneurs. Text: "Stop juggling tasks. Get your ducks in a row."
Expected AI response: a NON-literal translation ("stop juggling tasks — get your affairs in order" instead of the literal line about ducks); the upbeat tone and informal address preserved; in the adaptation list — "get your ducks in a row → get your affairs in order: an English idiom, the literal duck version doesn't work in Russian", plus a note anywhere ambiguity remains. Terms (SaaS features) left in their accepted form.
Variations
- Two versions. "Give me two translations: one closer to the letter and one a freer localization — I'll pick by platform."
- Reverse check. "Translate it back into the source language so I can see whether the meaning drifted" (back-translation).
- Glossary. Attach a list of terms "translate this way / leave untranslated" — for consistency across a large project.
Pro tips
- Context decides quality: the same text is translated differently for an ad and for a contract. Always give the text type, the tone and where it will be used — without that you get "correct but dead".
- Ask for the adaptation list (idioms, ambiguities) — it isn't filler, it's your review checklist: those are exactly the spots where machine translation most often misses the meaning or the tone.
- For legal, medical and other high-stakes texts, an AI translation is a draft: a specialist translator signs off on the final. Don't pour sensitive documents with personal data into a public AI — anonymize them or use a local model.