Zentrix Agent Studio supports multiple Indian languages and offers natural-sounding voices for each. Configure your agent's language and voice in Step 3 (Personalize) of the setup wizard.
| Language | Code | Notes |
|---|---|---|
| English | en | Default. Best accuracy for English-speaking callers. |
| Hindi | hi | Handles both pure Hindi and Hindi-English code-switching (Hinglish). |
| Marathi | mr | Marathi language transcription. |
| Auto-detect | auto | Agent responds in whichever language the caller speaks. Uses Hindi-mode transcription. |
Auto-detect mode is designed for businesses whose callers may speak English, Hindi, or a mix of both. Here is how it works:
Tip: Auto-detect works best for Hindi-English scenarios. If your callers speak primarily Marathi, set the language to mr explicitly for better transcription accuracy.
For businesses that want the agent to explicitly offer a language choice at the start of each call, you can include language selection in your agent's greeting message:
"Hello, thank you for calling Acme Corp. For English, please continue speaking
in English. Hindi mein baat karne ke liye, Hindi mein bolein."The auto-detect mode will then respond in whichever language the caller uses going forward.
Zentrix Agent Studio uses AI-synthesized voices across both the phone and web widget platforms. The available voices differ slightly between platforms:
Phone Calls
| Voice | Gender | Language | Voice ID |
|---|---|---|---|
| Nila | Female | Indian English / Hindi | Optimized for Indian accent |
| Sam | Male | English | Natural male voice |
| Rachel | Female | English | Natural female voice |
Web Widget
| Voice | Gender | Language |
|---|---|---|
| Samad | Male | Indian English |
| Monika | Female | Indian English |
| Amritanshu | Male | Indian English |
| Adrian | Male | American English |
| Dorothy | Female | American English |
The system selects a voice based on your agent's persona name:
The inference covers a wide range of common Indian and Western names. You can always override the selection by setting an explicit voice ID.
Speech-to-text transcription is handled by:
Our speech recognition engine provides primary transcription for all languages, with high accuracy for Indian-accented English and Hindi. It supports 16kHz sampling rate with linear16 encoding and 250ms endpointing for responsive real-time transcription.
The transcription runs in streaming mode, meaning the agent begins processing the caller's words as they speak -- it does not wait for the caller to finish their entire sentence. This enables natural, responsive conversations.
In the setup wizard, Step 3 (Personalize) provides:
Changes to language and voice are synced to both the phone and web widget platforms when you save.
Our text-to-speech voices use an advanced synthesis model, which provides:
Tip: For the most natural-sounding calls with Indian callers, use the Nila or Samad voices. They are specifically tuned for Indian English pronunciation and rhythm.
A Product by BRTNeura Technology LLP
Last updated: 2026-03-05