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

# The voice studio

> Where your voices live — which voice speaks for you in each language, trying a voice out, fine-tuning it, and tidying up.

**Where:** Proxies ▸ Voices

The voice studio is where you manage your voices once they exist. Making a new one is covered in
[Creating your voice](/docs/guides/presenters/your-voice.md).

![The voice studio](https://mintcdn.com/latentcast/IQao64bAvXRWz47P/images/guides/voice-studio.png?fit=max\&auto=format\&n=IQao64bAvXRWz47P\&q=85\&s=a2a489105748d9aecfc7a57ad071dbc7)

The voice studio: which voice speaks for you in each language on top, and every voice available to you below.

The page has three parts, top to bottom:

1. **My preferred voice collection**: which voice speaks for you in each language.
2. **Voices**: every voice available to you, yours and the ready-made ones.
3. **Pronunciation**: your dictionaries for names and terms, covered in
   [Pronunciation dictionaries](/docs/guides/names/pronunciation-dictionaries.md).

## Which voice speaks which language

Your **voice collection** (Which voice speaks for you in each language. A story picks from it automatically.) has one row per language. Each row shows the voice that
speaks for you in that language, with a ▶ button to hear it. The title counts how many languages are
set, for example *6 of 9 set*.

A story picks the voice for its language from this collection automatically, so you rarely think
about it. It matters when a story is made in a language that has no voice: that row shows **no voice
set**, and production stops and asks for one rather than speaking the language with the wrong voice.
If you are cast for Dutch or French work, set a voice for that language before you accept the
request.

To change a row, press ✎ (or **Set…** on an empty row) and pick a voice. Only voices that cover that
language are offered. If none do, add one first: when you [create a voice](/docs/guides/presenters/your-voice.md)
you choose its language, and the recording script comes in that language.

When you are cast, what gets bound to that outreach is your collection as it stands at that moment.
That is why changing it later does not disturb a campaign already running.

## Your voices

Every voice has a card. It shows the name, what kind of voice it is, the languages it covers, and a
★ when it is in your collection. The kinds:

| Label | What it is |
| - | - |
| **INSTANT CLONE** | Made from one short recording of you |
| **PRO CLONE** | Trained on longer recordings of you |
| **SYSTEM** | A ready-made voice provided by LatentCast |
| **TUNED** | Your own copy of a ready-made voice, with your settings |
| **CUSTOM** | A voice you imported by its ID |

A professional clone shows **training** while it is being made. That takes up to three hours, and the
card updates by itself when it is done.

Use the search box and the **Type**, **Language**, **Profile** and **Style** filters to find a voice in
a long list, and switch between the card view and a compact list.

## Trying a voice out

Press ▶ on a card, or **Preview & settings** in its ⋮ menu. Pick a language to hear a sample sentence,
or type your own and press **Generate & play**. Use a real sentence from your own outreach rather than
the sample: your customer's company name is a better test than a generic phrase.

If a name comes out wrong, that is fixable and it is not your voice's fault. See
[Pronunciation dictionaries](/docs/guides/names/pronunciation-dictionaries.md), or tell whoever runs your
campaigns.

A professional clone can come out of training as more than one version (**V1**, **V2**…). Listen to
each in the preview and commit the one you like best.

## Fine-tuning a voice

The same dialog has two tabs:

* **Settings**: how the voice speaks. What you can change depends on the engine: usually its
  **speed**, and depending on the engine how steady or expressive it is. Small changes go a long way.
  Try it with **Generate & play** before you press **Save settings**.
* **Languages**: which languages the voice covers. A voice needs at least one. Only tick a language
  if the voice really speaks it well; the collection uses this list to decide which voices it offers.

A ready-made voice cannot be changed itself. Saving settings on one makes your own **tuned** copy,
which appears in your list next to it.

## Other actions

The ⋮ menu on a card also has:

* **Set as preferred…**: put this voice in your collection for one or more of the languages it
  covers. The dialog shows which voice it would replace in each.
* **Rename**: or click the name on the card. Ready-made voices keep their name.
* **Delete**: removes one of your own voices; ready-made ones cannot be deleted. It cannot be undone.
  If a story still uses the voice, you are told to change that story's voice first.

An administrator also sees **Change speech profile**, which decides whose speech-provider account the
voice uses.

## Next

> **[Giving consent](/docs/guides/presenters/giving-consent.md)**
>
> With an avatar and a voice, agree to the request.

> **[Creating your voice](/docs/guides/presenters/your-voice.md)**
>
> Add another voice, for example for a new language.


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