How the AI text translator works
Modern AI translation does not swap words one by one from a dictionary. A large language model reads the whole passage, works out what it means and then writes the same meaning in another language. That is why the result usually sounds like a person wrote it, with natural word order and idioms, instead of a word-for-word copy.
Automatic source language detection
You do not need to say which language your text is in. The model recognises it from the text itself, so you only choose the language you want. This also works when a message mixes two languages.
Hinglish, the language many Indians actually type
Millions of people in India write Hindi in English letters on their phones: "kal meeting 5 baje hai", "aapka order aa gaya". Dictionary-style translators often stumble on this, because the words are not in any Hindi or English dictionary in that spelling. A language model reads it like a person does, so "time pe aa jana" becomes "please be on time". It also works the other way: choose Hinglish to turn English into the casual Roman Hindi people use in chats.
24 target languages and a choice of tone
This page translates into English, Hindi, Hinglish, Gujarati, Marathi, Tamil, Telugu, Bengali, Kannada, Punjabi and Urdu, and into Spanish, French, German, Portuguese, Italian, Arabic, Chinese, Japanese, Korean, Russian, Turkish, Dutch and Indonesian. The tone option (auto, formal, casual, friendly or professional) matters in languages where politeness is built into the grammar, such as German, French, Hindi, Japanese and Korean, so a business email does not sound like a text to a friend.
What translates well and what needs care
Everyday messages, emails, product descriptions, reviews, support replies and general articles usually translate well. Idioms, humour, legal clauses, medical instructions and brand names are riskier, because a small change in meaning can matter a lot. For anything important, have a native speaker check the result.
AI translation compared with a dictionary-style translator
Because the model understands context, it can pick the right meaning of a word with several senses, keep a consistent register and rephrase a sentence that does not work literally in the other language. The trade-off is that it can occasionally add or drop a detail, so compare the translation with the original when accuracy matters.
Translation inside your own product
The translation API supports several target languages in one request, glossaries and protected terms (so a product name is never translated), which suits support tools, e-commerce catalogues and multilingual websites. The free quota applies on every plan.