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

# Voice Selection

> Choose or create the voice your AI agent will use on calls.

The voice you choose significantly affects how contacts perceive your AI agent. Atllas provides a library of pre-built voices and the ability to clone your own.

## Voice Library

The voice library contains a curated set of voices powered by leading TTS providers (including ElevenLabs). Each voice has a name, gender, and sample you can play before selecting.

To choose a voice:

1. Browse or search the library
2. Click a voice to hear a preview
3. Click **Select** to use it for your campaign

## Voice Settings

Once a voice is selected, you can fine-tune its delivery:

| Setting              | Description                                                                                                                                                                                    |
| -------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Speech Rate**      | How fast the AI speaks. Default is 1.0. The available range depends on the selected voice's provider — for ElevenLabs voices, the range is 0.7–1.2                                             |
| **Stability**        | (ElevenLabs voices only, surfaced when Speech Rate is non-default) How consistent the voice sounds across sentences. Higher stability = more uniform; lower = more expressive variation        |
| **Similarity Boost** | (ElevenLabs voices only, surfaced when Speech Rate is non-default) How closely the output matches the original voice sample. Higher values are more accurate but can introduce audio artifacts |

For most campaigns, the defaults work well. Adjust if the voice sounds robotic or inconsistent on test calls.

## Voice Cloning

Voice cloning lets you create a custom voice from an audio recording — useful if you want the AI to sound like a specific person on your team.

### Requirements

* At least 30 seconds of clean audio (no background noise, music, or echo)
* Supported formats: MP3, WAV, M4A
* The speaker should use a natural, conversational tone

### Creating a Cloned Voice

1. Go to **Settings → Voices**
2. Click **Create Voice Clone**
3. Upload your audio file and give the voice a name
4. The clone is processed in a few minutes and then available in the voice library

Your account includes up to **2 custom voice slots**. Deleting a voice clone frees the slot for a new one.

## Language Support

Voices are available in multiple languages. When selecting a voice, filter by language to find options that match your target audience. If you select a non-English voice, the AI will respond in that language throughout the call.

<Warning>
  Ensure your call script and call details are written in the same language as the voice you select. Mixed-language configurations may produce inconsistent results.
</Warning>