> ## Documentation Index
> Fetch the complete documentation index at: https://docs.asteragents.com/llms.txt
> Use this file to discover all available pages before exploring further.

# ElevenLabs Text to Speech

> Generate natural-sounding spoken audio from text using ElevenLabs voices

## What it does

Converts text into spoken audio using [ElevenLabs](https://elevenlabs.io) voice synthesis. The generated MP3 is attached to the conversation so it can be played back, downloaded, or handed to other tools (for example, transcribed back via [elevenlabs\_speech\_to\_text](/tools/elevenlabs_speech_to_text)).

## Key features

* Curated enum of 10 voices spanning gender, accent, and tone — pick by name, no voice IDs required
* Escape hatch for custom or cloned voices via raw `voice_id`
* Latest `eleven_v3` model by default; opt into `eleven_multilingual_v2`, `eleven_flash_v2_5`, or `eleven_turbo_v2_5` for specialized tradeoffs
* Configurable output format (MP3 at multiple bitrates, PCM, µ-law)
* Output file is attached to the thread for downstream use

## Parameters

| Parameter       | Type   | Required | Description                                                                                                                                                        |
| --------------- | ------ | -------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `text`          | string | Yes      | The text to speak (up to 10,000 characters)                                                                                                                        |
| `voice`         | enum   | No       | One of `rachel`, `sarah`, `jessica`, `charlotte`, `lily`, `george`, `brian`, `daniel`, `will`, `charlie`. Defaults to `rachel`.                                    |
| `voice_id`      | string | No       | Raw ElevenLabs voice ID (e.g. a cloned voice). Takes precedence over `voice` if set.                                                                               |
| `model_id`      | enum   | No       | `eleven_v3` (default, most expressive), `eleven_multilingual_v2` (29 languages), `eleven_flash_v2_5` (\~75ms latency), `eleven_turbo_v2_5` (speed/quality balance) |
| `output_format` | enum   | No       | `mp3_44100_128` (default), `mp3_44100_192`, `mp3_22050_32`, `pcm_16000`, `pcm_22050`, `pcm_24000`, `pcm_44100`, `ulaw_8000`                                        |
| `filename`      | string | No       | Filename for the generated file. Extension is added automatically. Defaults to `tts_<timestamp>.mp3`.                                                              |

## Voice reference

| `voice`     | Gender | Accent     | Best for                    |
| ----------- | ------ | ---------- | --------------------------- |
| `rachel`    | F      | American   | Calm narration, explainers  |
| `sarah`     | F      | American   | Soft professional voiceover |
| `jessica`   | F      | American   | Expressive, upbeat reads    |
| `charlotte` | F      | British    | Sultry narration            |
| `lily`      | F      | British    | Warm, conversational        |
| `george`    | M      | British    | Warm narrator, storytelling |
| `brian`     | M      | American   | Deep, authoritative         |
| `daniel`    | M      | British    | News/authoritative          |
| `will`      | M      | American   | Chill, conversational       |
| `charlie`   | M      | Australian | Casual                      |

## Common use cases

### Generate a short voiceover

```
text: "Welcome to the onboarding flow. Let's get you set up."
voice: "rachel"
```

### Use a specific model for low-latency playback

```
text: "Your order has shipped."
voice: "will"
model_id: "eleven_flash_v2_5"
```

### Use a custom cloned voice

```
text: "This is our branded voice."
voice_id: "<your_custom_voice_id>"
```

## Response

Returns a `generated_files` attachment with the MP3, plus metadata about the voice, model, output format, and file size. The attachment is immediately available to downstream tools in the same conversation.

## Setup

No per-user setup. ElevenLabs is configured at the platform level — just enable the tool on your agent in **Control Hub > Edit Agent** under the **Audio** section.