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

# Voices

> Choosing a catalog voice, cloning your own from an audio sample, and matching a voice to an agent's name.

Manage voices at **Voices** in the Command Center. There are three places to look: **Explore**, a catalog of ready-made voices you can filter and preview; **My voices**, the voices your organization has cloned; and **Defaults**, a short list of presets tuned for common intake moods.

## Choosing a catalog voice

The Explore tab lists each catalog voice with a short description, its language, accent, and style, along with a **Preview** button that plays a real sample. Filter by style (conversational, professional), accent, or language to narrow the list. The Defaults tab has three starting points: a structured, concise voice for straightforward intake, a softer-paced voice for distressed or uncertain callers, and a more formal, direct voice for high-value or high-complexity calls.

## Cloning your own voice

On **My voices**, select **Create voice** (or **Upload sample**) to clone a voice from an audio sample.

<Steps>
  <Step title="Name the voice">
    Give it a name your team will recognize, such as an attorney's name.
  </Step>

  <Step title="Upload a sample">
    Upload a WAV or MP3 file. A clean clip of a few seconds up to a few minutes works; longer recordings are automatically trimmed to a short opening clip before cloning.
  </Step>

  <Step title="Pick the source channel">
    If the recording has two people on it, for example a recorded call, choose which channel to use: the whole mix, or the left or right channel alone.
  </Step>

  <Step title="Detect speakers (optional)">
    For a two-person recording, use **Detect speakers** to split it and preview each one, then select the speaker you want to clone instead of the whole clip.
  </Step>

  <Step title="Upload & clone">
    Submit the sample. The voice appears in **My voices** with a status while it processes.
  </Step>
</Steps>

Each cloned voice shows a status, **Processing**, **Ready**, **Couldn't build**, or **Removed**, and once ready, a quality badge (**Excellent**, **Good**, or **Low**) so you can judge whether it's usable before assigning it to an agent. **Preview** plays a real synthesized sample of that voice, not a generic system voice, so what you hear is what callers will hear.

## Matching the agent's name to its voice

An agent's name (what it calls itself on a call, set in the agent editor) and its voice are configured separately, so it's worth checking that the two sound like the same person. A name that reads as male paired with a voice that reads as female, or the reverse, can read as inconsistent to a caller. Pick the name and voice together when you set up an agent, and revisit the pairing if you change one without the other.