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

# Audio models

> ElevenLabs V3, Turbo v2.5, Multilingual v2, and Gemini 3.1 Flash TTS — covering high-quality speech, fast iteration TTS, multilingual coverage, and voice cloning for Studio and MCP workflows.

Elaris supports text-to-speech (TTS), dialogue generation, and voice cloning through ElevenLabs and Google models. Credits are deducted per generation action.

<Tabs>
  <Tab title="ElevenLabs">
    **Industry-leading TTS with direct and fal routing paths.**

    | Model           | Notes                                         |
    | --------------- | --------------------------------------------- |
    | ElevenLabs V3   | Default high-quality TTS; natural prosody     |
    | Turbo v2.5      | Faster generation; lower latency              |
    | Multilingual v2 | Broader language support for global campaigns |

    ElevenLabs models support **voice clone** inputs — upload a reference audio clip and the model matches the voice timbre and style. Voice cloning carries a higher credit cost than standard TTS.

    <Tip>
      For Studio narration, **ElevenLabs V3** is the recommended default. Use **Turbo v2.5** when iterating quickly on script drafts where speed matters more than fine prosody.
    </Tip>
  </Tab>

  <Tab title="Google">
    **High-quality TTS from Google's Gemini family.**

    | Model                | Notes                                 |
    | -------------------- | ------------------------------------- |
    | Gemini 3.1 Flash TTS | Fast and high-quality alternative TTS |

    Gemini TTS is a strong alternative when you need tight integration with other Gemini-based generation steps in a pipeline, or when ElevenLabs voice availability is a constraint.
  </Tab>
</Tabs>

## What audio generation modes are available?

| Mode                   | Description                                | Relative cost   |
| ---------------------- | ------------------------------------------ | --------------- |
| **TTS (single voice)** | Convert a script to speech with one voice  | Base cost       |
| **Dialogue**           | Multi-speaker conversation generation      | Higher than TTS |
| **Voice clone**        | Match a target voice from a reference clip | Higher than TTS |

## How do I use audio in MCP?

The `generate_audio` MCP tool accepts the following parameters:

```json theme={"dark"}
{
  "model": "elevenlabs-v3",
  "voice": "your_voice_id",
  "language": "en",
  "stability": 0.75
}
```

Discovery tools (`list_models` with `kind: "audio"`) are **free**. Each `generate_audio` call deducts credits based on model and output length.

<CardGroup cols={2}>
  <Card title="How pricing works" icon="calculator" href="/docs/pricing/how-pricing-works">
    Understand credit deductions for TTS, dialogue, and voice clone modes.
  </Card>

  <Card title="Models overview" icon="layers" href="/docs/models/overview">
    Browse the full catalog including image and video models.
  </Card>
</CardGroup>