Indian language voice agents on Bolna

Bolna voice agents run in 20+ languages. This section has a guide for each Indian language below: which transcribers (speech-to-text) and voices (text-to-speech) support it, the language code to send, how to start calls in it, and a minimal create-agent API example. Each language needs a transcriber and a voice that both support it. The providers listed here are the ones whose own pages in these docs confirm the language. A provider not listed may still work, but place a test call first.

---

​

Language guides

---

​

Providers by language

“Auto-detect” means Bolna can detect the language automatically mid-call on a multilingual agent. For the provider notes behind this table, including ElevenLabs model coverage and Assamese, see Indian languages by provider.

---

​

What every language guide covers

• Language code. Send the plain code, such as ta, as transcriber.language and in provider_config.language, not a locale such as ta-IN. Odia is od everywhere except Maya, which takes or.
• Starting a call in a language. On a multilingual agent, pass "agent_data": { "language": "<code>" } on POST /call, or an agent_language column in a batch CSV. See Start a call in a chosen language.
• Switching mid-call. See How language switching works and Auto-switch languages for system messages.
• Prompts. Write each language’s prompt in its native script. See Writing prompts in non-English languages.
• Recognition. Bias the transcriber toward names and terms with keywords and context.

For every language and code you can add to an agent, including non-Indian languages, see Supported languages.

---

​

Next steps