voice_model

Force Vocalizer to use an older TTS technology when synthesizing the selected voice model.

Value
  • full_encryptf8 (8 Khz)
  • full_vssq5f22 (22Khz)
  • XPremium (22Khz)
  • XPremium-nb (Narrow band–8Khz)
  • XPremium-high (22Khz)
  • XPremium-high-nb (8Khz)

Default

(determined by the installed voice)

How to set

Shown in precedence order (see Rules of parameter precedence):

Set in a VoiceXML application, set as switts.voice_model using a <property> element.

Set in a voice browser, with switts.voice_model using the Vendor-Specific-Parameters header in a SPEAK method.

Set in the session xml file for the application. Configure the <speechsynth> element located under <sessionparameters>, <speechserver>.

Note: This parameter can also be set on the Vocalizer host. However, that setting is overridden by the methods listed here.

Usage

Seldom used.

By default, Vocalizer uses the best quality model available. Use this parameter when you install multiple voice models and want to lock in an older voice technology to be compatible with applications built under Vocalizer 5.0 (or earlier) that were heavily tuned with tuned ActivePrompt databases.

When you download voices, you'll find separate folders for the voice models:

Nuance Network folder Voice models
Standard XPremium
Enhanced XPremium-high

The XPremium-high and XPremium-high-nb voice models produce higher quality speech, but use more disk space, CPU, and memory. Though XPremium-high voices are packaged at 22 kHz, you can configure them to produce output at any sample rate.

Note: To reduce latency when using XPremium-high voices, enable early_emission_allowed.

For more, see Vocalizer languages and voices.