# Voice and calls

> Let visitors speak their questions and hear answers read aloud, or call your bot from the chat, and what each costs.

Voice settings are on your bot's **Model** step, under **Voice**.

## Choose how visitors talk

1. Turn on **Visitors can speak their questions**. A microphone appears in the chat.
2. Optionally, turn on **Read answers aloud**.
3. Under **Whose voice**, choose one:

| Option | What visitors get | What it costs |
|---|---|---|
| AI voice | Natural voices in many languages, from OneInfer. | Your OneInfer credits: about $0.004 for an answer read aloud, and a small fraction of a cent for a spoken question, on the cheapest models. |
| The visitor's browser | The voice and the languages their device has. Some browsers, such as Firefox, can't listen. | Free. |
| A call with the bot | Visitors talk with the bot as on the phone, and it answers as they speak, in any language. They can talk over it. | Your OneInfer credits, from about 2¢ to about 10¢ a minute depending on the model, plus Aura's fee for the conversation. |

Recordings aren't kept. What each side says shows in the chat as text.

## A voice conversation

With **AI voice** or **The visitor's browser**, the microphone starts a hands-free conversation. Visitors talk, and what they say is sent as their message when they pause. With **Read answers aloud** on, answers are read from their first sentence, and visitors can talk over them. Each answer also gets a play button.

For AI voice, you can choose the **Speaking model**, the **Listening model**, the **Voice** and its speed, and use **Hear this voice** to try it. The cheapest model is chosen unless you pick another.

## A call with the bot

With **A call with the bot**, the microphone starts a call. Choose the **Call model**: the cheapest is the default.

| Setting | Default | What it does |
|---|---|---|
| Hang up after | 30 seconds of quiet | From 10 to 120 seconds. |
| Longest call | 5 minutes | From 1 to 30 minutes. |

A call is many times the cost of a typed conversation: a 3-minute call on a by-the-minute model is about 26¢, against about 2.5¢ for a typed chat. So calls hang up on quiet and have a time limit. A call also ends when your team takes the chat over, when the visitor says goodbye, or when they ask for a person, which brings in your team.

Each visit can make up to 10 calls an hour, and each call counts as one answer towards the bot's daily limit.

> [!NOTE]
> Your privacy policy needs a line about voice. The Privacy text on **Go live** adds it when voice is on. See [Privacy](https://aura.oneinfer.ai/docs/privacy).