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

# Your own TTS

> Bring your own voice provider - any voice from your library, 2p/min off

With **your own TTS**, the agent speaks with a voice from **your own** voice-provider account - any voice in your library, cloned voices included - and the per-minute price drops by **\$0.02**, because Nixflex no longer pays for the speech.

Supported providers: **ElevenLabs** and **Cartesia**. Each is a hand-built streaming integration - voice is the most latency-sensitive part of a call, so this list is curated rather than open.

<Note>
  The connection is **tested before it is saved**: Nixflex runs one real synthesis against your account with the exact voice ID you enter. A wrong voice ID or a dead key is refused at save with the provider's reason - never discovered on a live call.
</Note>

## How it works

<Steps>
  <Step title="Find your voice ID">
    In your provider's console, pick a voice and copy its ID (ElevenLabs: the Voices page; Cartesia: the voice's UUID).
  </Step>

  <Step title="Connect it">
    In **Settings → Your own TTS**, or via the API below: provider, voice ID, and your provider API key.
  </Step>

  <Step title="Your voice speaks every phone call">
    Applies account-wide. Your provider bills you for the characters; the Nixflex rate drops by \$0.02/min on calls your voice actually served.
  </Step>
</Steps>

<Note>
  **Phone calls only.** Web calls use high-quality 48kHz audio and keep your agent's standard voice - the telephone integrations produce 8kHz phone audio, and Nixflex never downgrades web audio.
</Note>

## If your voice fails

* The **standard voice takes over instantly** for the rest of the call - callers never hear silence.
* The call's **Events** tab records a `PROVIDER_FALLBACK` event with the reason.
* **That call bills at the standard TTS rate** - the discount applies only to calls your voice actually carried.

## Setting it up via the API

```bash theme={null}
curl -X PUT https://api.nixflex.com/v1/account/tts \
  -H "Authorization: Bearer KEY_ID:KEY_SECRET" \
  -H "Content-Type: application/json" \
  -d '{
    "provider": "elevenlabs",
    "voice_id": "21m00Tcm4TlvDq8ikWAM",
    "api_key": "sk_..."
  }'
```

| Field      | Required | Notes                                             |
| ---------- | -------- | ------------------------------------------------- |
| `provider` | Yes      | `elevenlabs` or `cartesia`                        |
| `voice_id` | Yes      | Exactly as shown in your provider's console       |
| `api_key`  | Yes      | **Write-only** - stored encrypted, never returned |

Read the configuration (the key is never included):

```bash theme={null}
curl https://api.nixflex.com/v1/account/tts \
  -H "Authorization: Bearer KEY_ID:KEY_SECRET"
```

Disconnect (calls return to the standard voice and rate from the next call):

```bash theme={null}
curl -X DELETE https://api.nixflex.com/v1/account/tts \
  -H "Authorization: Bearer KEY_ID:KEY_SECRET"
```

## Stacking the discounts

Bring everything and the rate compounds:

| You bring          | Rate            |
| ------------------ | --------------- |
| Nothing (standard) | \$0.08/min      |
| Your LLM           | \$0.065/min     |
| Your TTS           | \$0.06/min      |
| LLM + TTS          | **\$0.045/min** |

Premium features (web calls, live monitoring) add \$0.01/min only on calls that actually use them - the same actual-use law as the discounts.

## Related

* [Your own LLM](/advanced/your-own-llm) - bring your own model too
* [Your own storage](/advanced/your-own-storage) - recordings in your own bucket