# Turn a script into speech with Voice Generator

> Use the Nodespell Voice Generator Creative Tool to turn a written script into spoken audio, choosing a voice, language and delivery style.

Source: https://docs.nodespell.com/creative-tools/voice-generator/ · Last updated: 2026-11-10

Voice Generator is the Nodespell Creative Tool for text-to-speech. You write a script, pick a speech model and a voice, and get back an audio file of the script being spoken. Voice Generator lists only speech models; for music or sound effects, use the other audio tools.

## When to use Voice Generator

| You want…                                           | Use                                                                       |
| --------------------------------------------------- | ------------------------------------------------------------------------- |
| A voiceover, narration or line of dialogue as audio | **Voice Generator**                                                       |
| To browse speech, music and effects models together | [Audio Generator](https://docs.nodespell.com/creative-tools/audio-generator/)                       |
| A face that says the line                           | [Speak](https://docs.nodespell.com/creative-tools/speak/), optionally using Voice Generator's audio |
| A song with sung lyrics                             | [Music Generator](https://docs.nodespell.com/creative-tools/music-generator/)                       |

## Generate speech

1. Switch the workspace to **Create**. Select **Audio Generator** in the tool rail, open the tool switcher at the top of the form, and choose **Voice Generator**. It is also under **All tools**.
2. Check the model. The default is **ElevenLabs TTS V3**.
3. Type the script in the text field.
4. Choose a **Voice**, and adjust delivery settings such as speed or style if the model offers them.
5. Set **Generations** from 1 to 4 if you want several reads to choose from.
6. Select **Generate**.

## Voice Generator controls

| Control                             | What it does                                        | Options and defaults                                                       |
| ----------------------------------- | --------------------------------------------------- | -------------------------------------------------------------------------- |
| Tool switcher (title)               | Moves to another audio Creative Tool                | Audio Generator, Voice Generator, Music Generator, Sound Effects Generator |
| Model picker                        | Chooses a speech model                              | Default: ElevenLabs TTS V3                                                 |
| **Previous model** / **Next model** | Steps through speech models                         | —                                                                          |
| **More settings**                   | Less-used model settings                            | Collapsed                                                                  |
| **Generations**                     | Number of runs submitted together                   | 1 to 4, default 1                                                          |
| **Generate**                        | Adds or updates the model node and submits the runs | Shows the credit estimate                                                  |

## Recommended models for Voice Generator

The **Featured** collection lists ElevenLabs TTS V3 (default) and Gemini 3.1 Flash TTS. **All models** also includes ElevenLabs TTS Turbo V2.5, Kling Text To Speech, Kokoro 8.2m, Minimax Speech 02 HD, Minimax Speech 02 Turbo, Qwen3 TTS, Resemble AI Chatterbox and Resemble AI Chatterbox Pro.

### ElevenLabs TTS V3 settings

| Setting                          | What it does                                                         | Default         |
| -------------------------------- | -------------------------------------------------------------------- | --------------- |
| **Text**                         | The script to speak                                                  | Empty           |
| **Voice**                        | Preset voice, listed by gender and accent, such as "F - US - Rachel" | F - US - Rachel |
| **Speed**                        | 0.7 to 1.2; below 1 is slower                                        | 1               |
| **Stability**                    | 0 to 1; lower is more expressive, higher is more consistent          | 0.5             |
| **Similarity Boost**             | 0 to 1; how closely to follow the original voice                     | 0.75            |
| **Style**                        | 0 to 1; exaggerates the voice's speaking style                       | 0               |
| **Previous Text**, **Next Text** | Surrounding lines that help continuity when you split a long script  | Empty           |

### Gemini 3.1 Flash TTS settings

Gemini 3.1 Flash TTS reads inline delivery tags in the prompt, such as `[laughing]`, `[whispering]` or `[short pause]`. **Style Instructions** describe tone, pace and accent without being spoken. For dialogue, add **Speakers** entries whose **Speaker Alias** matches the line prefix in the script, for example `Host:`, and give each its own **Voice**. **Language Code** defaults to English (US).

## What changes with the model?

- **Voices:** every model has its own voice list; voices do not carry over when you switch.
- **Languages:** some models offer a language setting; check the model's description for supported languages.
- **Reference audio:** a few models, such as Qwen3 TTS and Resemble AI Chatterbox, accept an audio sample to guide the voice.
- **Multiple speakers:** Gemini 3.1 Flash TTS supports multi-speaker scripts; most models read one voice.

When you switch models, Nodespell carries over the settings it can map and tells you which could not be carried over.

## Where does the audio go?

**Generate** places the speech model node in the current graph level and runs it. The audio appears in **Results** in the Create form and stays in the node's history; see [Takes](https://docs.nodespell.com/graph/takes/). Generated audio also appears under **Generated** and **Audio** in the [Asset Manager](https://docs.nodespell.com/assets/asset-manager/), ready to use in [Speak](https://docs.nodespell.com/creative-tools/speak/) or the [Timeline](https://docs.nodespell.com/edit/timeline/).

## Common problems

- **Words are mispronounced.** Spell them phonetically in the script, or try another voice or model.
- **The read sounds flat or erratic.** On ElevenLabs TTS V3, lower **Stability** for more emotion or raise it for a steadier read.
- **Long scripts drift between parts.** Split the script and use **Previous Text** and **Next Text** so each part knows its neighbors.

## Frequently asked questions

### What is the default Voice Generator model?

The default is ElevenLabs TTS V3. It has a long list of preset voices and controls for speed, stability, similarity and style.

### Can Voice Generator create a dialogue with two voices?

Yes, with Gemini 3.1 Flash TTS. Add a speaker entry for each alias used in the script and choose a voice for each.

### Can I clone my own voice?

Some models accept reference audio to guide the voice, such as Qwen3 TTS and Resemble AI Chatterbox. Use audio you have the right to use.

### How do I make a character say the line on screen?

Generate the audio here, then use it as the audio input in [Speak](https://docs.nodespell.com/creative-tools/speak/).

## Related

- [Audio Generator](https://docs.nodespell.com/creative-tools/audio-generator/) — all audio models in one picker
- [Speak](https://docs.nodespell.com/creative-tools/speak/) — put the voice on a face
- [Make a talking video](https://docs.nodespell.com/recipes/talking-video/) — script to voice to video
- [Prompts and saved text](https://docs.nodespell.com/graph/prompts-and-saved-text/) — reuse scripts
- [Choose a model](https://docs.nodespell.com/creative-tools/choose-a-model/) — compare speech models