ACD and attendant console

Announcements and voices

Generate announcements with local speech synthesis, choose a voice, optionally connect cloud voices.

EasyUC · Version 4.8.7

In EasyUC: Calling Features → Calls → Calling Features → Voice settings / Call Flow

What it is for

Announcements are created from text. EasyUC generates them with a speech synthesis that runs on your own server (Piper). The text does not leave the network. The generated audio file is cached, so the same text is computed only once. The announcement is sent to the caller as G.711 at 8 kHz.

You set language and voice once on the script. Every speech block in the call flow and every announcement of a queue can deviate from that.

Optionally, you can connect an ElevenLabs account. The blocks then additionally offer a cloud engine with the voices of that account. The media server calls the provider during playback and needs internet access for this.

Prerequisites

Licence module acd or teams.
Permission: ROLE_APP_ACD_WORKSPACE and ROLE_APP_ACD_SCRIPTS to view, additionally ROLE_APP_ACD_SCRIPTS_EDIT to change. Administrators with ROLE_APP_ADMIN, ROLE_OWNER, ROLE_ADMIN or ROLE_SUPER_ADMIN have all rights of the area.
A running media server. It contains the local voices.
The delivered state includes German voices: one female and one male default voice and one further male voice for explicit selection. For languages without their own voice model, the service uses a simpler fallback synthesis (eSpeak).
For cloud voices only: an ElevenLabs API key and internet access for the media server.

Step by step: Set the voice for a script

Open Calling Features → Calls → Calling Features and click Edit on the script.
Select the Language.
Under Voice gender, select Female or Male.
Under Voice, select the specific voice.
Click Save.

Step by step: Write an announcement in the call flow and listen to it

On the script, open the Call Flow and click the block Play AI Text, Menu (DTMF) or Queue.
Under Voice engine, leave the selection Local (Piper · offline).
If required, adjust Language, Voice gender and Voice for this block.
Enter the text. Write phone numbers and abbreviations the way they are to be spoken.
Click Preview voice. The announcement is generated and played in the browser.
Click Apply and then Save flow.

Step by step: Connect cloud voices (optional)

Open Calling Features → Calls → Calling Features.
Click Voice settings at the top. The button shows the current engine, Piper or ElevenLabs.
Under ElevenLabs API key, enter the key.
Under Model, select the speech model.
Click Save.
Open a call flow, select ElevenLabs (Cloud) under Voice engine in the speech block and then the ElevenLabs voice.

Fields and options

Fields and options
Field / action	Meaning and use
Voice engine	Local (Piper · offline): synthesis on your own server. ElevenLabs (Cloud): only selectable with a stored key, otherwise shown with the addition no API key.
Language	Language of the text. Determines the voices offered.
Voice gender	Female or Male. Without an explicitly selected voice, the service uses the default voice of the gender.
Voice	Voices of the media server for language and gender.
ElevenLabs API key	Displayed masked. Leaving it empty keeps the stored key, a new entry replaces it.
Model	`eleven_multilingual_v2` (best quality), `eleven_turbo_v2_5` (faster, cheaper), `eleven_flash_v2_5` (fastest, lowest latency).
Credits	At the top of the call flow: remaining and total characters of the ElevenLabs account, queried live.
Sound file (mp3/wav)	Upload field in the block Play Soundfile and for the hold music. For the state of playback, see chapter "Call flow: all blocks".

Check the result and narrow down errors

Preview voice reports Erzeugt & im Cache gespeichert (generated and stored in the cache) and plays the announcement.
Call the script number and listen to the announcement on the phone. The telephone quality (8 kHz) sounds duller than the preview in the browser.
No voice for this language/gender: There is no voice model for the combination. Change the gender or the language.
No voices for this language. The ElevenLabs account has no voice for the language.
Vorhören fehlgeschlagen (HTTP …). (preview failed) The speech service is not reachable. Check the media server.
Abspielen im Browser nicht möglich. (playback in the browser not possible) The announcement was generated, the browser blocked the playback. It still works on the phone.
The cloud announcement stays silent: The media server does not reach the provider, or the credits are used up.