# Sarvam

> Built for Indian languages. Every listening setting for Sarvam, and how to set it.

Source: https://voxio.in/docs/voice-listening-sarvam · voxioagent 2.0.0 · Index: https://voxio.in/docs/llms.txt

Service: `sarvam`. A new setup starts from:

```json
{
  "service": "sarvam",
  "model": "saaras:v3",
  "language": "hi-IN",
  "mode": "transcribe"
}
```

## Set it

On the flow's **Voice** tab, choose **Listening**, then **Sarvam**. Or from code:

```js
await voxio.flows.update(flowKey, {
  stt: {"service":"sarvam","model":"saaras:v3","language":"hi-IN","mode":"transcribe"},
});
```

## Settings

Settings marked “under More settings” are behind **More settings** in the dashboard.

- `model` (string): **Model**. Which speech model listens. Hover an option to compare. Default: `"saaras:v3"`.
- `language` (string): **Language**. The language callers speak, or detect it. Default: `"hi-IN"`.
- `mode` (string): **Output**. How the transcript is written. Default: `"transcribe"`.
- `keyterm` (string[]): **Boost words** (under More settings). Words the agent should never mishear.
- `vad-signals` (boolean): **Speech-start events** (under More settings). Signal the moment the caller starts speaking. Default: `true`.
- `high-vad-sensitivity` (boolean): **Hear quiet speakers** (under More settings). Pick up soft voices; may also pick up background noise. Default: `false`.
- `positive-speech-threshold` (number): **Speech threshold** (under More settings). How confident it must be that someone is speaking. From 0 to 1. Default: `0.7`.
- `negative-speech-threshold` (number): **Silence threshold** (under More settings). How confident it must be that speech has stopped. From 0 to 1. Default: `0.45`.

## Choices

### Model choices

| Value | Name | Notes |
| --- | --- | --- |
| `saaras:v4` | Saaras v4 | Latest. Newest Sarvam model, better across Indian languages. |
| `saaras:v3` | Saaras v3 | Reliable for Hindi and English calls. |

### Language choices

| Value | Name | Notes |
| --- | --- | --- |
| `unknown` | Detect automatically |  |
| `en-IN` | English (India) |  |
| `hi-IN` | Hindi |  |
| `bn-IN` | Bengali |  |
| `gu-IN` | Gujarati |  |
| `kn-IN` | Kannada |  |
| `ml-IN` | Malayalam |  |
| `mr-IN` | Marathi |  |
| `od-IN` | Odia |  |
| `pa-IN` | Punjabi |  |
| `ta-IN` | Tamil |  |
| `te-IN` | Telugu |  |

### Output choices

| Value | Name | Notes |
| --- | --- | --- |
| `transcribe` | Transcribe | In the spoken language |
| `translate` | Translate to English |  |
| `codemix` | Code-mixed | Hinglish-style |
| `translit` | Transliterate | In Latin script |
| `verbatim` | Verbatim | Every filler word |

## Next steps

- [Listening](https://voxio.in/docs/voice-listening.md): compare the other providers.
- [Turn-taking](https://voxio.in/docs/turn-taking.md): when the agent answers.