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

# Speech and voices

> Create narration with voice discovery, expressive delivery and export settings.

<Frame>
  <img src="https://chat.youniverse.ai/__l5e/assets-v1/982b9812-45cc-4e3a-be1d-6432d2bfceea/docs-studio-speech.jpg" alt="Glass studio microphone, open script and audio wave" />
</Frame>

Open **AI Studio → Speech** to turn written text into narration. **Settings** configures the next generation; **History** opens previous speech.

## Choose a voice

The voice picker has **Explore** and **My Voices**, with language, accent, category and gender filters. Preview voices before committing to a long script. Voice choice and delivery settings work together; the same words can sound very different across voices.

## Choose a model

| Model | Intended use |
| - | - |
| Flash v2.5 | Fast speech generation. |
| Multilingual v2 | Natural multilingual narration and longer-form speech. |
| v3 (expressive) | Emotional delivery and supported inline audio tags. |

Model limits vary. Use the character count and selected model's limit shown in the workspace rather than treating every model as interchangeable.

## Delivery controls

| Control | What it changes |
| - | - |
| Speed | How quickly the voice speaks. |
| Stability | Delivery consistency versus variation. |
| Similarity | Adherence to the selected voice's character. |
| Style Exaggeration | Strength of the voice's stylistic delivery where supported. |
| Speaker boost | Voice similarity enhancement where supported. |
| Language | The language used for synthesis. |
| Output format | Available PCM, MP3 or Opus export settings. |

Some controls depend on the model. Do not expect a disabled setting to affect the result.

<Frame>
  <img src="https://chat.youniverse.ai/__l5e/assets-v1/d3f0d9c7-fbe0-4fd6-8c25-155cfa442c22/docs-studio-voices.jpg" alt="Three glass microphones with distinct voice waveforms" />
</Frame>

## Expressive scripts

With a compatible expressive model, tags such as `[laughs]`, `[whispers]` and `[sighs]` can guide delivery. Place them close to the phrase they should affect. Punctuation and short paragraphs also help make a script easier to speak naturally.

## Reuse speech from History

Search Speech History for a previous generation. **Use** restores its text and voice for another narration; **Download** keeps the audio, and **Delete** removes an unwanted history item.

## Generate, listen and export

<Steps>
  <Step title="Write the script">Read it aloud, expand ambiguous abbreviations and check names.</Step>
  <Step title="Set voice and delivery">Preview the voice, choose the model and adjust supported settings.</Step>
  <Step title="Review the estimate and generate">Longer scripts and model choice affect usage.</Step>
  <Step title="Listen and download">Check pronunciation and pacing, then download the result in the selected supported format or use narration in [Lip sync](/studio/lip-sync).</Step>
</Steps>

<AccordionGroup>
  <Accordion title="Is this the same as Iris speaking in chat?">No. Speech is a script-to-audio production workspace. [Voice chat](/chat/voice) is a live conversation.</Accordion>
  <Accordion title="Can I return to an earlier narration?">Open Speech History to find previous generations.</Accordion>
  <Accordion title="Does the ZDR badge also cover a lip-sync video?">No. Read the badge for each operation separately; Imagine animation is not covered by Speech's badge.</Accordion>
</AccordionGroup>

<Info>Speech requires Pro or Elite. The Speech workspace displays its own privacy badge; always review that badge and the credit estimate before generating.</Info>


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