# Voice Design

> Create a new voice from a written description, with control over model, loudness, guidance and seed. Get an audio preview and a voice ID you can keep and reuse.

Source: https://nodaro.ai/docs/nodes/audio/voice-design

The **Voice Design** node creates a new voice from a written description. You describe the voice, give it a text to speak, and the node returns an audio preview of the new voice and the voice's ID. Unlike [Voice Remix](https://nodaro.ai/docs/nodes/audio/voice-remix), it gives you control over the model, loudness, guidance, seed and quality, so you can refine a voice and reproduce it.

- Found in: Audio › Voices
- Output: audio
- API type: `voice-design`

## When to use it
- You want a brand voice with settings you can reproduce across a project.
- You want character voices with specific qualities, such as age, accent, tone and delivery.
- You want to refine a voice step by step, one setting at a time.
- You want a voice ID to keep, instead of only a preview.

## Quick start
### Add the node

Press Tab on the canvas and choose **Audio › Voices › Voice Design**.

### Describe the voice

Open the settings panel and write the **Voice Description**, for example: "A warm, deep male voice with a British accent and a calm delivery."

### Write the preview text

Write 100 to 1,000 characters under **Preview Text**. The new voice speaks this text in the preview. You can also wire a text node into the **Prompt** input to supply it.

### Run it

Click **Run** on the node. The preview appears on the node with a player, and the voice ID appears under it. Click the voice ID to copy it.

Workflow: Two voice pickers describe the voice, a Text node supplies the preview lines, and Voice Design creates the voice.

- Voice Character → Voice Design (audio style)
- Voice Delivery → Voice Design (audio style)
- Text → Voice Design (prompt)

## Inputs and outputs
| Input | Accepts | What it does |
| --- | --- | --- |
| **Prompt** | Text nodes, such as Text and Prompt | The preview text. When connected, it is used instead of the Preview Text field. |
| **Audio style** | Voice pickers, such as Voice Character and Voice Delivery | Adds the pickers' wording to the voice description. The wording is shortened when needed, so that the whole description stays within 1,000 characters. |

| Output | What it carries |
| --- | --- |
| **Audio** | The audio preview: the new voice speaking the preview text. |
| **Voice ID** | The ID of the new voice. The settings panel also shows it under **Generated Voice ID**. |

## Settings
| Setting | What it does |
| --- | --- |
| **Voice Description** | The voice in plain words: age, gender, accent, tone, pace and emotion. |
| **Preview Text (100-1000 chars)** | The text the new voice speaks in the preview, from 100 to 1,000 characters. With ElevenLabs v3, type `[` to insert an audio tag. |
| **Model** | **ElevenLabs v3 (recommended)**, the default, or **ElevenLabs Multilingual v2**. |
| **Loudness** | From -1 (quiet) to 1 (loud). The default is 0, the model's normal level. |
| **Guidance Scale** | From 0 (creative) to 100 (strict). The default is 5. How literally the model follows the description. |
| **Seed (optional)** | A fixed number that makes a voice reproducible. Leave it empty for a new voice on every run. |
| **Quality (optional)** | A higher value gives better quality and less variety. Leave it empty for the model's default. |
| **Enhance audio quality** | Cleans up the preview audio after generation. Off by default. |
| **Pre & post text** | Text that is always added before and after the description. It is hidden from people who use your workflow as an app. See [Prompt pre and post text](https://nodaro.ai/docs/concepts/prompt-pre-post-text). |

## Models
| Model | Maker | Modes | Credits | Details |
| --- | --- | --- | --- | --- |
| [ElevenLabs Voice Design](https://nodaro.ai/docs/models/audio/elevenlabs-voice-design) | ElevenLabs | Voice design | 50 | Design a synthetic voice from a description (no reference clip needed). |

Use **ElevenLabs v3**, the default, for the best quality and the widest language support. Switch to **ElevenLabs Multilingual v2** only when v3 gives an unexpected result for a specific voice.

## Write a good description

- **Name concrete qualities.** "Gravelly baritone, speaks slowly, sounds like a late-night radio host" works better than "a cool, interesting voice".
- **Cover the basics.** Age range, gender, accent, tone, pace and emotional quality.
- **Match the preview text to the final use.** Write the preview text in the style of the final content, so you hear the voice as it will be used.

## Keep and reuse a voice

- **Fix the seed.** When a run gives a voice you like, write down its **Seed**, description and model. Changing any of them can change the voice.
- **Copy the voice ID.** After a run, the voice ID shows on the node and under **Generated Voice ID** in the settings panel. Click it on the node to copy it.
- **Balance several voices.** Use **Loudness** to bring voices you design to the same level.

## Voice Design or Voice Remix

| | Voice Design | [Voice Remix](https://nodaro.ai/docs/nodes/audio/voice-remix) |
| --- | --- | --- |
| **Controls** | Model, loudness, guidance, seed, quality, audio enhancement | Description and preview text only |
| **Outputs** | Audio preview and voice ID | Audio preview |
| **Best for** | A voice you will keep and reproduce | Fast exploration of voice ideas |

## Tips
- **Start at Guidance Scale 5.** Raise it toward 20 or 30 for a precise description. Lower it toward 1 or 2 for a freer, more natural voice. Very high values can reduce quality.
- **Leave Enhance audio quality off while you explore.** Turn it on for the final run, because it adds processing time.
- **Explore with Voice Remix first.** Try ideas quickly in [Voice Remix](https://nodaro.ai/docs/nodes/audio/voice-remix), then recreate the best one here with full control.

## From the API
`POST /v1/voice-design` takes `text` (the preview text), `voiceDescription` and the optional `model`, `loudness`, `guidanceScale`, `seed`, `quality` and `shouldEnhance`. The MCP tool is `voice_design`. See [Voice and media](https://nodaro.ai/docs/developers/api/voice-and-media).

## Frequently asked questions

### What is the difference between Voice Design and Voice Remix?

Both create a voice from a written description. Voice Design adds controls for model, loudness, guidance, seed and quality, and it also returns a voice ID. Voice Remix is the quick version, with only a description and a preview text.

### How long must the preview text be?

Between 100 and 1,000 characters. Write enough text to hear the voice across different sounds and sentence shapes, for example two or three natural sentences.

### How do I get the same voice again?

Set a Seed when you find a voice you like. The same seed, description and model should give the same voice again. Changing any setting can give a different voice, even with the same seed.

### What does Guidance Scale do?

It sets how strictly the model follows your description. The default is 5. Raise it toward 20 or 30 for a precise description, or lower it toward 1 or 2 for a more natural result.

### Can I describe the voice with pickers instead of words?

Yes. Wire a Voice Character picker, for age, gender, language, accent and timbre, or a Voice Delivery picker, for pace, emotion and archetype, into the Audio style input. Their wording is added to your description.