Give a bot a voice

Member·4 minutes to read

What this does for you

This page covers choosing a voice for a bot, calling it, hearing it speak, and talking to it instead of typing.

Before you start

  • The bot already exists, and you can reach its Settings pane — see Make a bot, and set it up.
  • Calling and speech need a voice route your organisation allows; if none is configured yet, an operator sets one up before any of this works.

Steps

  1. Open the bot's Settings and find the Voice row. <!-- shot: voice-01-phone.png | The Voice row in a bot's Settings, on a phone --> <!-- shot: voice-01-desktop.png | The Voice row in a bot's Settings, on a desktop -->
  2. Choose a voice from the list. Not set is first and is the default, with no preview button on it; every other voice is listed alphabetically, each with a you can press to hear a sample without leaving the list. Until your organisation has a voice route it allows, the list holds only Not set. <!-- shot: voice-02-phone.png | The voice list open, on a phone --> <!-- shot: voice-02-desktop.png | The voice list open, on a desktop -->
  3. Start a call from the bot's details panel, once it has a voice chosen. The call control stays off until both a voice is set on the bot and an administrator has accepted where the call audio is processed. <!-- shot: voice-03-phone.png | Starting a call, on a phone --> <!-- shot: voice-03-desktop.png | Starting a call, on a desktop -->
  4. Talk with the bot. If your organisation keeps voice data in the EU, the call takes turns: hold the button while you speak and release it when you're done. Otherwise, speak whenever you like and the bot answers as you go. <!-- shot: voice-04-phone.png | An open call, on a phone --> <!-- shot: voice-04-desktop.png | An open call, on a desktop -->
  5. End the call when you're done. A call also ends on its own after 30 minutes. <!-- shot: voice-05-phone.png | Ending a call, on a phone --> <!-- shot: voice-05-desktop.png | Ending a call, on a desktop -->
  6. Ask the bot to speak a passage back to you instead of typing it, for a digest or a summary you'd rather hear. The audio is always marked as synthetic, so you know it was generated rather than recorded. <!-- shot: voice-06-phone.png | A synthetic audio message in the transcript, on a phone --> <!-- shot: voice-06-desktop.png | A synthetic audio message in the transcript, on a desktop -->
  7. Hold the composer's microphone button to dictate a message instead of typing it, when your organisation has a transcription route configured. <!-- shot: voice-07-phone.png | Dictating into the composer, on a phone --> <!-- shot: voice-07-desktop.png | Dictating into the composer, on a desktop -->

Image generation in a bot's answers, and a bot sitting in on a video meeting, are not available yet.

A worked example

Asking a bot to speak a digest instead of writing it out:

Read me this morning's digest instead of posting it as text.

What can go wrong

What you seeWhyWhat to do
No voice route is configured for this organisation yet, so there is no voice to choose. An operator sets one up.Nobody has configured a voice route your organisation is allowed to use.Ask an operator to add one.
Calling is switched off for your organisation. An administrator can turn it on in settings.Calling itself is switched off.Ask an administrator to turn it on.
Before the first call, an administrator has to accept where the audio is processed.Nobody has accepted the residency of call audio yet.Ask an administrator to open voice settings and accept it.
This Bot has no voice yet. Choose one in Settings and the call control turns on.The bot's own Voice field is still Not set.Choose a voice in the bot's Settings.
That is too long to turn into one audio file. Ask for a shorter passage.The passage you asked the bot to speak is over the length a single audio file can hold.Ask for a shorter passage, or split it into more than one.

See also

Last verified against build 9e6bdf2.