# TTS Settings

Set how fast the Assistant speaks with `tts.speed` in an [Advanced Settings](/docs/advanced-settings) resource.

## Overview

The platform translates `speed` into the selected TTS provider's native range. Increasing the value makes speech faster across supported integrations, but the audible change varies by provider.

## Data model

| Field   | Type           | Default                   | Description                                                            |
| ------- | -------------- | ------------------------- | ---------------------------------------------------------------------- |
| `speed` | number \| null | `null` (normal pace) | Speech speed from `-1` (slowest) to `1` (fastest). `0` is the configured normal pace |

Values outside `-1` to `1` are rejected with a `400` response. When `speed` is `null` or `0`, the platform uses its configured normal pace for that provider. Some integrations send that normal value explicitly; Gradium uses a normal `padding_bonus` of `-1.5`.

Set `speed` to `0.25` to make speech slightly faster. The request path is relative to the Data API server URL in the [API reference](/api), which includes `/data_next`:

```http title="HTTP Request"
PATCH /ai/advancedSettings/{your-advanced-settings-id}
Authorization: Bearer {your-api-key}
Content-Type: application/json
```
```json title="Payload"
{
  "tts": {
    "speed": 0.25
  }
}
```

***

## How the value is translated

The platform interpolates linearly between the slowest, default and fastest speeds that each provider handles well. A `speed` of `0.5` lands halfway between the provider's default and its fastest usable value, and `-0.5` halfway between the default and its slowest. The platform uses provider-specific bounds. Some are narrower than the provider's native range, while others use the full supported range.

| Model                    | `speed`                  |
| ------------------------ | ------------------------ |
| Cartesia Sonic-3.5       | Yes                      |
| Cartesia Sonic-3         | Yes                      |
| ElevenLabs Flash v2.5    | Yes                      |
| Google Chirp-3           | Yes                      |
| Gradium Default          | Yes                      |
| Voxtral Mini TTS         | No, the value is ignored |
| Inworld TTS-1            | No, the value is ignored |
| Gemini 2.5 flash preview | No, the value is ignored |

On a provider that exposes no speed control, the value is accepted and saved but has no effect on the call. On an Assistant using only an **STS (Speech-to-Speech)** model, there is no TTS stage, so the `tts` object does not apply. An STS Assistant configured with a separate TTS model uses that TTS integration and its speed support.

:::info{title="Note"}

Speed is a per-Assistant setting, not a per-sentence one. If a specific passage needs to be read slowly, such as a confirmation code or a phone number, ask for it in the instructions instead of lowering the global speed.

:::

<br />

## What's next

- [Advanced Settings](/docs/advanced-settings)
- [STT Settings](/docs/stt-settings)
- [VAD Settings](/docs/vad-settings)
