# Voice

Your agent answers the phone. What a call sounds like, what you configure, and where each setting lives.

Source: https://docs.omazy.ai/how-to/voice/

import Figure from '../../../../components/Figure.astro'
import { CardGrid, LinkCard } from '@astrojs/starlight/components'

Voice gives your agent a phone number. Someone calls it, the agent picks up,
and the two of them have a conversation out loud.

It is the same agent that answers on your website. Same brief, same knowledge,
same catalog. What changes is the medium, and the medium changes what good
looks like: a caller cannot skim, cannot scroll back, and hears silence as a
dropped line rather than as thinking.

Most of the voice settings exist to deal with that one difference.

## What a call sounds like

<Figure
  label="The stages of a single phone call"
  caption="The two highlighted stages are spoken from phrases you author. They are the parts a caller judges the fastest, and the only parts that can be ready before the call arrives."
>
<svg viewBox="0 0 700 150" xmlns="http://www.w3.org/2000/svg">
  <text x="0" y="14" class="d-eyebrow">ONE CALL, START TO FINISH</text>

  <rect x="0" y="28" width="120" height="46" rx="8" class="d-box" />
  <text x="14" y="48" class="d-label">Dials</text>
  <text x="14" y="64" class="d-sub">your number</text>

  <path d="M124 51 L134 51" class="d-arrow" />
  <polygon points="140,51 133,47.5 133,54.5" class="d-arrow-head" />

  <rect x="140" y="28" width="120" height="46" rx="8" class="d-box-accent d-pulse" />
  <text x="154" y="48" class="d-label">Greeting</text>
  <text x="154" y="64" class="d-sub">no wait</text>

  <path d="M264 51 L274 51" class="d-arrow" />
  <polygon points="280,51 273,47.5 273,54.5" class="d-arrow-head" />

  <rect x="280" y="28" width="120" height="46" rx="8" class="d-box" />
  <text x="294" y="48" class="d-label">Listens</text>
  <text x="294" y="64" class="d-sub">caller speaks</text>

  <path d="M404 51 L414 51" class="d-arrow" />
  <polygon points="420,51 413,47.5 413,54.5" class="d-arrow-head" />

  <rect x="420" y="28" width="120" height="46" rx="8" class="d-box-accent d-pulse" />
  <text x="434" y="48" class="d-label">Acknowledges</text>
  <text x="434" y="64" class="d-sub">while it thinks</text>

  <path d="M544 51 L554 51" class="d-arrow" />
  <polygon points="560,51 553,47.5 553,54.5" class="d-arrow-head" />

  <rect x="560" y="28" width="120" height="46" rx="8" class="d-box" />
  <text x="574" y="48" class="d-label">Answers</text>
  <text x="574" y="64" class="d-sub">then loops</text>

  <path d="M140 88 L140 96 L540 96 L540 88" class="d-arrow" />
  <text x="248" y="116" class="d-accent-text">spoken from phrases you author, per language</text>
</svg>
</Figure>

The loop repeats until someone hangs up. Everything after the greeting is the
agent listening, thinking and speaking, over and over, fast enough that it
feels like a conversation.

## What you configure

| Setting | Where | What it decides |
|---|---|---|
| **Voice** | Agent, Saved Responses, Voice | Who the caller hears, per language |
| **Spoken phrases** | Agent, Saved Responses, Voice | The exact words for greetings, acknowledgements, nudges and farewells |
| **Languages** | Agent, Saved Responses, Voice | Which languages the agent speaks, and which one leads |
| **Pacing** | Agent, Saved Responses, Voice | How the line handles delay and echo |
| **Number** | Voice, Numbers | The phone number that reaches this agent |
| **Call log** | Voice | Recordings, transcripts and what was said |

The brief, the knowledge and the catalog are shared with every other channel.
You do not write a separate agent for the phone. You add a phone voice to the
agent you already have.

## Start here

<CardGrid>
  <LinkCard
    title="Set up a voice agent"
    href="/how-to/voice/setup/"
    description="The order the settings depend on each other in, and why doing it out of order gives you a silent line."
  />
  <LinkCard
    title="Write the spoken phrases"
    href="/how-to/voice/phrases/"
    description="Greetings, acknowledgements, idle nudges and farewells, per language."
  />
  <LinkCard
    title="Change the voice"
    href="/how-to/voice/changing-voice/"
    description="What changes for the first few callers, and the weekly limit that exists because of it."
  />
  <LinkCard
    title="Call quality"
    href="/how-to/voice/call-quality/"
    description="Pacing, echo and the two settings that decide whether the agent talks over people."
  />
  <LinkCard
    title="The call log"
    href="/how-to/voice/call-log/"
    description="Recordings, transcripts, filters, and turning a good answer into a saved response."
  />
  <LinkCard
    title="Troubleshooting"
    href="/how-to/voice/troubleshooting/"
    description="Wrong voice, talking over the caller, dead air, and the call that ends in an apology."
  />
  <LinkCard
    title="Recovering lost calls"
    href="/how-to/voice/recovering-lost-calls/"
    description="A call that happened but never reached the log: how to spot it, and how the record is rebuilt."
  />
</CardGrid>