Gujarati, Tamil and Telugu
Live now. Send Gujarati, Tamil or Telugu text and you get audio in that language from the same speech endpoint, with the same voices and at the same price. You don’t add a language parameter or any other field, and you don’t change any code, because the model reads the script you send.- Send text in its own script: Gujarati in ગુજરાતી, Tamil in தமிழ், Telugu in తెలుగు. That is what the model reads.
- English words stay in Latin script, mixed in mid-sentence, exactly the way they already work in Hindi.
- Speech recognition is live too. Mira STT transcribes the same languages
with
gu-IN,ta-INandte-IN. See STT languages.
Kannada, Malayalam, Bengali and Marathi
Beta. All six voices speak Kannada, Malayalam, Bengali and Marathi. Kannada, Malayalam and Bengali work the same way as Gujarati, Tamil and Telugu: send the text in its own script and you get audio in that language from the same speech endpoint, at the same price, with no language parameter. Marathi is written in Devanagari like Hindi, so it is recognised by its words (आहे, नाही, तुमच्या) rather than its script. An agent whoselanguage is kn, ml, bn or mr speaks it on phone and web calls.
- Send text in its own script: Kannada in ಕನ್ನಡ, Malayalam in മലയാളം, Bengali in বাংলা, Marathi in Devanagari (मराठी). English words stay in Latin script.
- Marathi is recognised by its words. A line that is clearly Marathi is
read as Marathi; a very short one with no Marathi word in it (“धन्यवाद”) is
read as Hindi. Add
"language": "mr"on the speech endpoint to make every line Marathi. See Marathi in the quickstart. - Where it works today: phone and web calls,
POST /v1/audio/speech, the streaming socket and the playground in the console. Realtime sessions come later. - Speech recognition is in preview. Mira STT accepts
kn-IN,ml-IN,bn-INandmr-IN. See STT languages. - Beta means the voices are ready to try and we are still tuning them on real calls. Tell us what you hear.
Why these six
Gujarati, Tamil and Telugu are live, and Kannada, Marathi and Bengali are in beta. Together these six languages are the mother tongue of 429 million people, roughly every third Indian. For most of them, Hindi is no substitute. The census records how many speakers of each language also speak Hindi, and for four of the six the answer is almost none.
357 million people cannot be called in Hindi at all, so a voice agent that
speaks only Hindi cannot reach them. Only 1.5% of Tamil speakers report
speaking Hindi; for Telugu it is 5.7%, for Kannada 4.7% and for Bengali 8.6%.
With Hindi and these six together, an agent can speak the mother tongue of
about four in five Indians.
Speaker counts are first-language speakers from the 2011 Census of India;
“also speak Hindi” is Hindi reported as a second language (table C-17). These
are the most recent official figures. The next census is due in 2027.
What support means
- Same endpoint, same request shape. The
speech endpoint does not change. You send text in the
language’s own script, code-switched with English the way real conversation
is, and get audio back. Marathi shares its script with Hindi, so it is
recognised by its words, and the optional
"language": "mr"makes it certain. - The same voices. Every live language uses the voices in the voice gallery. You keep the voice you chose and change only the text.
- Speech recognition alongside. Gujarati, Tamil and Telugu are live in speech to text as well as in text to speech. Marathi, Kannada, Malayalam and Bengali are in preview there today.