One Greeting, Four Different Tones
Every morning, the dream team's bots post a short greeting on Instagram and Slack. The same line gets translated into Korean, Traditional Chinese, Simplified Chinese, and English before publishing. One day Bella lined up four screenshots side by side and noticed something odd.
「The Korean sounds like a friend talking, but the English reads like a corporate email.」
The same greeting carried a different temperature in each language. The Traditional Chinese version was overly polite and felt distant, while the Simplified Chinese version swung the other way, sounding too casual and light.
The Problem Wasn't Translation, It Was the Missing Standard
When Windi checked, the issue wasn't translation quality. Each language had been prompted separately, so each version was finished in its own session with its own feel. The real cause was that there was no shared tone reference across languages.
The Fix, Building a Tone Dictionary
Windi built a shared tone dictionary for all four languages. It set a standard like 「friendly but still respectful」, then mapped out sentence endings and phrases in each language that matched that standard. From then on, every translation prompt referenced this dictionary first.
The Result
Starting the next day, the four language greetings carried a similar warmth. That tone dictionary later carried over to card news captions and article translations too, becoming the shared voice standard for all of the dream team's content. Comparing four screenshots ended up reshaping an entire content system.