> For the complete documentation index, see [llms.txt](https://docs.borndigital.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.borndigital.ai/digital-agent/voice.md).

# Voice agents

How a Digital Agent handles a phone call or a voice conversation in the bubble, and where you set up speech recognition, the voice, listening and timing.

The same flow can answer in the chat and on the phone. In a voice conversation, a few more parts come into play:

1. **Speech synthesis (TTS)** reads out the **Speech** text of your nodes with the voice chosen in the deployment.
2. When the agent finishes speaking, **speech recognition (STT)** starts listening and turns what the customer says into text.
3. The flow evaluates the text in an [Answer node](https://docs.borndigital.ai/digital-agent/conversation-flow/nodes-explained/answer-node.md) – or a signal when something else happened: the customer was silent (`no_input`), spoke for too long (`shut_up`) or could not be understood (`no_match`).

## Where you set it up

| What                                                   | Where                                                                                                                                                                                                                                   |
| ------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Speech recognition and voice providers, the voice      | [Deploy Project](https://docs.borndigital.ai/digital-agent/workspace/deploy-project.md#speech-to-text) – **Speech to text** and **Voice**. See [Speech providers](https://docs.borndigital.ai/digital-agent/voice/speech-providers.md). |
| What the agent says                                    | The **Speech** text of Message and Answer nodes – see [Customizing speech synthesis](https://docs.borndigital.ai/for-advanced-users/conversation-design-tips/customizing-speech-synthesis.md).                                          |
| How long the agent listens, interruptions, keypad      | Answer node › **Configuration** › **Timing** and **Listening features**. See [Listening and timing](https://docs.borndigital.ai/digital-agent/voice/listening-and-timing.md).                                                           |
| Changes during the call                                | Flow variables such as `stt_language` or `no_input_timeout`. See [Voice variables](https://docs.borndigital.ai/digital-agent/voice/voice-variables.md).                                                                                 |
| Phone numbers, background music, black- and whitelists | [Deploy Project](https://docs.borndigital.ai/digital-agent/workspace/deploy-project.md#extras).                                                                                                                                         |
| Call recordings and transcripts                        | [Recordings](https://docs.borndigital.ai/digital-agent/advanced-functions/recordings.md) – audio exists only with **Record calls** on.                                                                                                  |

{% hint style="info" %}
A [Realtime agent](https://docs.borndigital.ai/realtime-agent/introduction.md) works differently: one speech-to-speech model listens and speaks at once, without a flow. Use it when natural, fast turn-taking matters more than a precisely scripted flow.
{% endhint %}

## Checklist for a good voicebot

* **Write for the ear** – short sentences, one question at the end of the message. See [Writing for voice](https://docs.borndigital.ai/for-advanced-users/conversation-design-tips/customizing-speech-synthesis/writing-for-voice.md).
* **Handle silence and misunderstanding** in every Answer node with a fallback – see [Fallback](https://docs.borndigital.ai/digital-agent/conversation-flow/nodes-explained/answer-node.md#fallback).
* **Set the timing per question** – a short timeout for yes/no, continuous transcription for open questions.
* **Check the latency** in the Deploy dialog – every provider and an AI node add to the time the caller waits.
* **Test with a real call** before you go live, in a quiet and in a noisy place.
