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

# ElevenLabs

> Connect ElevenLabs to SentiVue Talk

ElevenLabs provides the **voice** that speaks your agent's responses, and **transcription** of what the caller says. One connection covers both.

## What you need

| Field   | Required |
| ------- | -------- |
| API key | Yes      |

## Getting your API key

1. Sign in at [elevenlabs.io](https://elevenlabs.io/app/settings/api-keys).
2. Create a new key.
3. Copy it — the key is shown once.

<Warning>
  ElevenLabs keys are **permission-scoped**. A key created with only text-to-speech permission will not transcribe, and the failure appears at call time rather than when you save the connection.

  Grant both **Text to Speech** and **Speech to Text** if you want the connection to drive voice and transcription.
</Warning>

Keys can also carry a **character limit**. A key that hits its cap stops working mid-call, so either leave the limit unset or set it well above your expected volume.

<Note>
  For anything running in production, create the key under a **service account** rather than your own login. Personal keys are tied to your user and stop working if your access changes.
</Note>

## Adding the connection

1. Go to **Connections** and choose **ElevenLabs**.
2. Paste the API key.
3. Save.

SentiVue Talk validates the key as you save. Note that validation confirms the key works, not that it carries every permission you need — see the warning above.

## What you can select afterwards

Choose your models on [Model Configurations](/talk/configurations/inference-providers).

**Voice model** — start with the fastest and move up only if the voice is not good enough on a real call. See ElevenLabs' [models page](https://elevenlabs.io/docs/models).

**Voice** — pick from the dropdown, or enter a **voice ID** for one from your own library or a cloned voice. IDs look like `JBFqnCBsd6RMkjVDRZzb` and are shown against each voice in the ElevenLabs app.

**Language** — set it to match your callers. Transcription runs on Scribe — see the [speech-to-text guide](https://elevenlabs.io/docs/capabilities/speech-to-text).

## Using ElevenLabs for only one service

If you want ElevenLabs voices but a different transcriber, turn the transcription service off on the connection.

## Custom hosts

If you reach ElevenLabs through a proxy rather than `api.elevenlabs.io`, you can set a base URL per service on the connection.

## Troubleshooting

<AccordionGroup>
  <Accordion title="Voice works but transcription fails">
    The key is missing the speech-to-text permission. Create a new key with both permissions granted, or edit the existing key's permissions in the ElevenLabs app.
  </Accordion>

  <Accordion title="Calls worked and then stopped mid-month">
    Check whether the key has a character limit and has reached it. Keys stop working once capped.
  </Accordion>

  <Accordion title="A voice ID is rejected">
    Copy the exact ID from the ElevenLabs app rather than the voice name. Also confirm the voice is available to your account — a cloned voice belongs to the account that created it.
  </Accordion>
</AccordionGroup>