Any voice can speak any of the supported languages.
The way the current model works is that you don't select a specific language, you instead write in the language you want the AI to speak and the AI understands automatically. However, since there can be some overlap between different languages where they use similar words and vocabulary but with quite different pronunciations and accents, you should use a cloned voice that was cloned speaking the language with the correct accent.
The language is determined by the text, the accent and pronunciation is determined by the voice itself. It needs both of these contexts to function optimally.
We are looking into developing new technology to facilitate the selection of language, but the way the AI is built means that all of this is still in progress.