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

# Test Your Campaign

> Hear each voice before you pick it, and try your campaign in a chat panel on the review step before you launch any calls.

## Why Test First

A campaign only pays back when the agent sounds right and says the right things. Two checks in the wizard help you catch problems before you spend a call: a voice preview on the voice step, and a chat panel on the review step.

## Voice Preview

On the voice step of the campaign wizard, each built-in voice shows a play button. Click it to hear a short sample of that voice reading a fixed line, then decide whether to pick it. Click again to stop. Selecting a voice is a separate action, so previewing does not change your choice.

<Note>
  Voice previews are available for built-in voices. Cloned voices show no preview button since the sample comes from your own recording.
</Note>

### What To Listen For

* **Tone match.** Does the voice fit the persona your campaign describes?
* **Pace.** Fast voices sound urgent, slower voices sound patient. Match it to the call type.
* **Language fit.** A voice tagged for one language may still speak another, but the accent can carry over. Prefer a voice tagged for the language your leads actually use.

## Campaign Chat Test

On the review step, a chat panel lets you talk to your campaign in text. The first message is the campaign's opening line, exactly as it would be spoken at the start of a real call. Type a reply and the agent responds using your campaign's real instructions, first sentence, knowledge base, and tools.

The chat runs against the same production configuration your calls will use. If you change an instruction or attach a knowledge base, the next chat message reflects that change.

### What Chat Covers

* The agent's opening line, with any `{{lead_name}}` or similar placeholders filled in from the test lead.
* The agent's replies to your typed messages, driven by your instructions.
* Tool behavior, including when the agent decides to end the call.
* The same content checks the call pipeline runs after each reply, surfaced in the chat panel so you can see whether the reply repeats itself or drifts off-language.

### What Chat Does Not Cover

Chat is a text simulation. It cannot reproduce everything a real call includes:

* Audio behavior. Voice tone, pauses, and interruptions do not exist in text.
* Speech recognition. Real calls transcribe what the lead says, so noise and accents can change what the agent hears.
* Silence and echo detection. Signals that watch for a stalled call only fire during real audio.

Use chat to catch the obvious problems (wrong opening line, weak objection handling, missing knowledge) then place a real test call to confirm audio behavior.

<Note>
  The chat is stateless. Each chat session starts fresh. Nothing is billed and no call record is created.
</Note>

## Suggested Test Flow

<Steps>
  <Step title="1. Preview voices">
    On the voice step, play a few voices and pick the one that fits your campaign's persona and language.
  </Step>

  <Step title="2. Chat through the opening">
    On the review step, read the opening line. If it sounds off, edit the first sentence and chat again.
  </Step>

  <Step title="3. Push on objections">
    Type the objections your leads actually raise. Check the agent's answers against your knowledge base and instructions.
  </Step>

  <Step title="4. Try an end-of-call path">
    Ask the agent something that should end the call. Confirm the agent ends cleanly.
  </Step>

  <Step title="5. Place one real call">
    Once chat looks right, run a real test call to confirm audio behavior before starting the campaign.
  </Step>
</Steps>

## Next Steps

<CardGroup cols={2}>
  <Card title="Voice Library & Cloning" icon="microphone" href="/guides/custom-voices">
    Pick a built-in voice or clone your own.
  </Card>

  <Card title="Knowledge Bases" icon="database" href="/guides/knowledge-bases">
    Give your agent the facts it needs to answer questions.
  </Card>
</CardGroup>