We recently added IBM Watson™ Text-to-Speech to our list of speech synthesis engine options, expanding the number of voices you dynamically synthesize as part of phone and web calls. This update brings 27 new, unique voices across 11 different languages and 14 different Dialects. This new set of voices also includes 18 of IBM Watson’s newer, more natural sounding neural voices based Deep Learning technology.

With our direct integration with IBM Watson Text-to-Speech (TTS) you can access these voices without having to touch the Watson APIs or an IBM account. More voice options give you more opportunities to choose a voice that fits your unique brand and application requirements. IBM Watson TTS also includes unique SSML options for customizing the voice to your specific brand and customer needs.

With these recent additions, Voximplant now offers nearly 150 voices through leading speech engine providers like Google, Amazon, Yandex, and Tinkoff.

Check out the full list of IBM voice options available in Voximplant:

Code	Neural	Dialect	IBM Ver	Special features
ar_AR_Omar		Arabic	1
de_DE_Birgit	✓	German	3
de_DE_Dieter	✓	German	3
de_DE_Erika	✓	German	3
en_GB_Kate	✓	British English	3
en_US_Allison	✓	American English	3	✓
en_US_Emily	✓	American English	3
en_US_Henry	✓	American English	3
en_US_Kevin	✓	American English	3
en_US_Lisa	✓	American English	3	✓
en_US_Michael	✓	American English	3	✓
en_US_Olivia	✓	American English	3
es_ES_Enrique	✓	Castilian Spanish	3
es_ES_Laura	✓	Castilian Spanish	3
es_LA_Sofia	✓	Latin American	3
es_US_Sofia	✓	American Spanish	3
fr_FR_Renee	✓	French	3
it_IT_Francesca	✓	Italian	3
ja_JP_Emi	✓	Japanese	3
ko_KR_Youngmi		Korean	1
ko_KR_Yuna		Korean	1
nl_NL_Emma		Dutch	1
nl_NL_Liam		Dutch	1
pt_BR_Isabela		Brazilian Portugese	3
zh_CN_LiNa		Mainland China	1
zh_CN_WangWei		Mainland China	1
zh_CN_ZhangJing		Mainland China	1

Using IBM Watson Voices in Voximplant is Easy

To use IBM Watson voices in Voximplant, - simply choose one of the IBM voices from the VoiceList options in the language object within ttsOptions when using call.say, IVR prompt, and createTTSPlayer. The example below demonstrates how you’d use the neural version of IBM’s Allison voice, using the call.say function:

// Pick your voice here
let ttsVoice = VoiceList.IBM.Neural.en_US_Allison
 
function onCallConnected(e) {
 // Extract just the name portion from the voice without the prepended language code
 let ttsVoiceNameRegex = RegExp(/_([A-Z][a-z]{1,})/).exec(ttsVoice.voice)
 let ttsVoiceName = ttsVoiceNameRegex ? ttsVoiceNameRegex[1] : "unknown"
  
 let sayOptions = {"language": ttsVoice, "ttsOptions": { "rate": "slow"}};
 call.say(`<speak>Hi, I am <emphasis>${ttsVoiceName}</emphasis> from <say-as interpret-as="letters">${ttsVoice.provider}</say-as>. <break time="1s"> See <emphasis>Voximplant.com</emphasis> for more information on speech synthesis <break time="3s"></speak>`, sayOptions)
 
 call.addEventListener(CallEvents.PlaybackFinished, ()=>{
   VoxEngine.terminate()
 })
}
 
VoxEngine.addEventListener(AppEvents.CallAlerting, (e) => {
 call = e.call
 call.addEventListener(CallEvents.Connected, onCallConnected)
 call.addEventListener(CallEvents.Disconnected, VoxEngine.terminate)
 call.answer()
})

Accessing IBM Watson Voices in Voximplant Kit

IBM Watson voices are also available in Voximplant Kit. You will find IBM and the voices listed above in the Text to Speech, Interactive Menu, and Dialogflow Connector blocks.

Pricing

IBM Watson voices are priced at $25 USD per 1M chars for both standard and neural voices. Billing is quantized in units of 10 characters. See our pricing page for more details.

IBM Watson Voice Features

Neural voices

As noted earlier, IBM Watson offers a neural voice option based on recent advancements in Deep Neural Network (DNN) technology for voices in many languages. Select from VoiceList.IBM.Neural in VoxEngine or from the neural options in Voximplant Kit.. In most cases, you should use the neural voice option if it is available unless you are using one of the advanced SSML features noted below.

SSML

Voximplant passes along all SSML tags encoded in prompts sent to TTS engines, including IBM. Remember SSML support details vary by speech engine. See IBM’s SSML support page for details and limitations. Voximplant also has a generic HowTo on using SSML in VoxEngine.

Enhanced SSML

Besides some standard SSML features, IBM also offers some advanced speech modification options. Their Expressive SSML for VoiceList.IBM.en_US_Allison allows changing the speaking style to an good news, apologetic, or uncertain tone using a custom express-as tag - i.e. <express-as type="GoodNews">. Note this does not work on the neural version of this voice. See IBM’s Expressive SSML guide for details.

IBM also offers a <voice-transformation> element that can make larger changes to how a voice sounds. This feature can be used to create a voice that sounds very different from the base model, customized to your parameters. You can make a voice sound younger, breathier, or softer along with many other fine tuning options. See IBM’s documentation for details and their demo of this feature. Voice transformation is only available on the following voices:

VoiceList.IBM.en_US_Allison
VoiceList.IBM.en_US_Lisa
VoiceList.IBM.en_US_Michael

Ready to enhance your voice experience?

● Listen to sample voices from IBM here.

VoxImplant iOS SDK update - background mode support is now available