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

# Your own voices

> Give your agents a voice from your own ElevenLabs, Fish Audio or Cartesia account: a cloned voice, a brand voice, or any voice in their library.

Bring a voice from your own account with a text-to-speech provider: your cloned voice, your brand's voice, or any voice in the provider's library. You add an API key from that account, import its voices, and pick one on an agent's **Avatar & voice** tab. The agent speaks in that voice with your API key.

| Provider | API key from | Note |
| - | - | - |
| ElevenLabs | [elevenlabs.io/app/settings/api-keys](https://elevenlabs.io/app/settings/api-keys) | Needs a paid ElevenLabs plan: ElevenLabs blocks other services from using free-plan keys. |
| Fish Audio | [fish.audio/app/api-keys](https://fish.audio/app/api-keys) | Speech is charged to your Fish Audio API credit: add some before your first call. |
| Cartesia | [play.cartesia.ai/keys](https://play.cartesia.ai/keys) | |

## Add an API key

**1. Open Voices**

Click **Voices** in the sidebar, then **Import voices**. With no API key yet, this starts by adding one. To add another key later, click **Add API key** above your keys.

**2. Pick the provider and paste your key**

Pick **ElevenLabs**, **Fish Audio** or **Cartesia**, paste the API key from your account with it, and optionally give it a name. The key is checked with the provider before it's saved.

**3. Import voices**

The voices the key can use open next.

Your key is stored encrypted and is never shown again: the dashboard shows its last 4 characters. It's only used to list your voices, play their samples and speak in them during calls.

## Import voices

Click **Import voices** at the top of the page, or next to a key, to see the voices it can use. With more than one key, pick the key at the top of the list. Search them by name, click a voice's play button to hear its sample, and click **Import**. A voice the provider has no sample of, such as one you cloned there, says a short sentence made with your API key, which your provider charges like any other speech.

| Provider | Voices listed |
| - | - |
| ElevenLabs | The voices in **My Voices** in your ElevenLabs account. Add a voice from ElevenLabs' library to My Voices first. |
| Fish Audio | The voices you made in your Fish Audio account. |
| Cartesia | Your voices and Cartesia's library. |

To import a voice that isn't listed, such as one from Fish Audio's library, paste its voice ID from the provider's site in **Or import by voice ID**.

Imported voices are listed under **My voices** on the **Voices** page, where you can play, rename or delete them. Deleting one here leaves it in your account with the provider.

## Give an agent your voice

On the agent's **Avatar & voice** tab, switch the **Voice** panel to **My voices** and pick a voice. **Model** sets the provider's model that speaks it:

| Provider | Models (the default first) |
| - | - |
| ElevenLabs | Flash v2.5 (fastest), Multilingual v2, Eleven v3, Eleven v3 conversational, Eleven v4 turbo |
| Fish Audio | S2.1 Pro, S2 Pro, S1 |
| Cartesia | Sonic 3.6, Sonic 3.5, Sonic 3 |

**Speaking speed** works as with the other voices: each provider uses the closest speed it supports.

## Keep your voice working

* A key, or a voice, that an agent speaks with can't be deleted: give the agent another voice first. The error names the agents using it.
* If the provider stops accepting your key (it was revoked, or your plan or credit ran out), calls with your voice can't speak. Keep the key valid and your provider account in credit.
* To change a key, add the new key, import the voices again from it, switch your agents to them, then delete the old key.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.