37 languages, both directions.

You're reading this page in one of them.

Detected automatically

18 languages Taiwa recognises from what the caller says. Your agent doesn't have to know what they're hearing.

English · Deutsch · Français · Español · Italiano · Nederlands · Polski · Português · Русский · 日本語 · 한국어 · 中文 · Svenska · Türkçe · Čeština · Bahasa Indonesia · Română · Українська

Chosen by your agent

19 more that translate the same way once the call is running, but can't be detected from speech yet. Your agent picks the customer's language; left on auto-detect, the call isn't translated.

العربية · Български · Català · Dansk · Ελληνικά · Suomi · Galego · עברית · हिन्दी · Hrvatski · Magyar · Bahasa Melayu · Norsk bokmål · Slovenčina · Slovenščina · தமிழ் · ไทย · Filipino · Tiếng Việt

Regional variants

9 variants for the spoken translation, so a customer in São Paulo hears Brazilian Portuguese and one in Montréal hears Canadian French.

English (UK) · English (US) · Español (Latinoamérica) · Français (Canada) · Français (France) · Português (Brasil) · Português (Portugal) · 简体中文 · 繁體中文

Any of them to any other. Not just to and from English.

Every pair, both ways

Your agent's language and your customer's are set separately, so a Polish-speaking agent can take a Turkish call without English sitting in the middle of it. Both sides of the call are translated — the customer hears your agent, and your agent hears the customer. It isn't a one-way subtitle track.

How the translation sounds

Every call runs on DeepL Voice. The list above is DeepL's, and so is the voice your customer hears.

Close to the speaker's own voice

In 14 languages DeepL speaks the translation itself, and can make it sound close to the person speaking. The other languages are spoken in a preset voice. Nothing is kept as a voice profile.

Tone, emotion and pace

DeepL's newest voice model, Aure, carries tone, emotion, emphasis and pace into the translated speech, automatically, in 8 languages: English, German, Spanish, French, Italian, Japanese, Portuguese and Korean. In the others the translated voice is neutral.

Partners for the newer languages

For the newer languages DeepL works with speech partners of its own. The trust page names them and says what they do.

Powered by DeepL →

They are not all equally easy, and we'd rather say which

If your volume is mostly German or Japanese, expect the delay to run a little longer than French or Spanish — those languages put the verb at the end, so the meaning of a sentence arrives last. It's a property of the language, not the software, and human interpreters hit the same wall.

The best way to understand it is to try it on your own calls.

Why there's a delay →

Which ones are you missing?

Tell us the languages and the volume, and we'll tell you what it costs — or that it isn't worth it.