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

# Voice Settings

> Configure your agent's voice and speech

Voice settings control how your agent sounds. Choose from multiple voice providers, adjust speaking style, and fine-tune the audio experience.

## Voice Providers

Retone supports two premium TTS providers:

<Tabs>
  <Tab title="Cartesia (Recommended)">
    **Cartesia Sonic 3** offers:

    * Ultra-low latency (\< 100ms)
    * Emotional expressiveness
    * Natural prosody
    * Speed and stability controls

    Best for: Most use cases requiring natural conversation
  </Tab>

  <Tab title="ElevenLabs">
    **ElevenLabs** offers:

    * High-quality voices
    * Voice cloning capability
    * Wide language support

    Best for: Specific voice requirements or cloned voices
  </Tab>
</Tabs>

## Basic Settings

### Voice Selection

Choose from available voices:

1. Click the **Voice** dropdown
2. Browse available voices by provider
3. Preview voices before selecting
4. Click to apply

### Speed

Control speaking rate:

| Value | Effect                   |
| ----- | ------------------------ |
| 0.5   | Half speed (very slow)   |
| 0.8   | Slightly slower          |
| 1.0   | Normal speed             |
| 1.2   | Slightly faster          |
| 2.0   | Double speed (very fast) |

<Tip>
  For phone calls, 0.9-1.1 typically sounds most natural.
</Tip>

### Volume

Output volume level (0-100):

```yaml theme={null}
Volume: 80  # Recommended for clear audio
```

### Responsiveness

How quickly the agent starts speaking (1-10):

| Value | Effect                            |
| ----- | --------------------------------- |
| 1-3   | Slower, more deliberate responses |
| 4-6   | Balanced (recommended)            |
| 7-10  | Faster, may interrupt caller      |

## Emotional Expression (Cartesia)

Cartesia Sonic 3 supports emotional speech:

### Emotion Types

| Emotion       | When to Use                             |
| ------------- | --------------------------------------- |
| `neutral`     | Professional, business contexts         |
| `friendly`    | Customer service, general conversations |
| `excited`     | Sales, promotions, positive news        |
| `sympathetic` | Handling complaints, showing empathy    |
| `curious`     | Asking questions, gathering information |
| `confident`   | Assertions, recommendations             |

### Intensity

| Level    | Effect                         |
| -------- | ------------------------------ |
| `low`    | Subtle emotional undertone     |
| `medium` | Noticeable but not exaggerated |
| `high`   | Strong emotional expression    |

```yaml theme={null}
Emotion:
  Type: friendly
  Intensity: medium
```

## Back-channeling

Natural acknowledgment sounds during conversation:

```yaml theme={null}
Back-channeling:
  Enabled: true
  Frequency: 5        # How often (0-10)
  Words:
    - "mm-hmm"
    - "I see"
    - "right"
    - "okay"
    - "got it"
```

### When Back-channeling Triggers

* During caller's longer statements
* At natural pauses
* When caller shares information
* To indicate active listening

<Note>
  Higher frequency (7-10) sounds very engaged but may feel excessive. Lower (1-3) is more subtle.
</Note>

## Pre-response Phrases

Filler phrases before AI generates a response:

```yaml theme={null}
Pre-response:
  Enabled: true
  Phrases:
    - "Let me check that for you..."
    - "One moment please..."
    - "Good question..."
    - "I can help with that..."
```

### Benefits

* Reduces perceived latency
* Sounds more natural
* Gives AI processing time

### Conditional Pre-responses

Trigger different phrases based on context:

```yaml theme={null}
Conditions:
  - trigger: "complex_query"
    phrase: "Let me look into that..."
  - trigger: "simple_query"
    phrase: "Sure..."
  - trigger: "emotional"
    phrase: "I understand..."
```

## Laughter Injection

Add natural laughter to appropriate moments:

```yaml theme={null}
Laughter:
  Enabled: true
```

When enabled, the AI may add subtle laughter in response to humorous comments from callers.

## Pronunciation Overrides

Custom pronunciations for specific words:

```yaml theme={null}
Pronunciations:
  "Acme": "ACK-mee"
  "API": "A P I"
  "SQL": "sequel"
  "GIF": "jiff"
  "nginx": "engine-x"
```

### Adding Pronunciations

1. Go to Voice Settings
2. Click **Pronunciation Overrides**
3. Add word and phonetic spelling
4. Test with voice preview

## Auto-enabled Features

These settings are enabled by default:

| Feature                         | Description                |
| ------------------------------- | -------------------------- |
| **Allow interruptions**         | Caller can interrupt agent |
| **Speech normalization**        | Numbers read naturally     |
| **Noise filtering**             | Background noise reduced   |
| **Background speech detection** | Handles side conversations |

## Configuration Example

Complete voice configuration:

```yaml theme={null}
Voice Settings:
  Provider: Cartesia
  Voice: "Sonic 3 - Professional Male"
  Speed: 1.0
  Volume: 80
  Responsiveness: 5

  Emotion:
    Type: friendly
    Intensity: medium

  Back-channeling:
    Enabled: true
    Frequency: 5
    Words: ["mm-hmm", "I see", "right"]

  Pre-response:
    Enabled: true
    Phrases:
      - "Let me check..."
      - "One moment..."

  Pronunciations:
    "TechFlow": "TECH-flow"
    "API": "A P I"
```

## Testing Voice Settings

1. Open the **Test Call** panel
2. Have a conversation with your agent
3. Listen for:
   * Natural pacing
   * Clear pronunciation
   * Appropriate emotion
   * Smooth back-channeling
4. Adjust settings and retest

## Best Practices

<CardGroup cols={2}>
  <Card title="Match Brand Voice" icon="building">
    Choose a voice that fits your company's personality.
  </Card>

  <Card title="Test on Phone" icon="phone">
    Audio sounds different on phone vs. computer. Test both.
  </Card>

  <Card title="Moderate Back-channeling" icon="volume-low">
    Too much feels robotic; too little feels distant.
  </Card>

  <Card title="Add Key Pronunciations" icon="spell-check">
    Ensure company names, products, and jargon sound correct.
  </Card>
</CardGroup>

## Related

<CardGroup cols={2}>
  <Card title="System Prompt" icon="terminal" href="/agent-config/system-prompt">
    Define agent personality
  </Card>

  <Card title="Environment" icon="volume-high" href="/agent-config/environment">
    Background sounds and ambience
  </Card>

  <Card title="Test Call" icon="phone" href="/test-agent/test-call">
    Test voice settings
  </Card>
</CardGroup>