Sharvani
Sharvani

Reputation: 111

Add support for various accents in Directline speech for Azure chat bot

We are working on a requirement where we need to support various accents for azure chat bot. Currently we have Directline speech enabled for the speech as below.

(async function () {
    var speechServicesTokenRes = await fetch(
       'https://eastus.api.cognitive.microsoft.com/sts/v1.0/issuetoken',
        {
           method: 'POST',
           headers: {
               'Ocp-Apim-Subscription-Key': '***************'
           }
        });

    if (speechServicesTokenRes.status === 200) {
        authorizationToken = await speechServicesTokenRes.text();

        var webSpeechPonyfillFactory = await window.WebChat.createCognitiveServicesSpeechServicesPonyfillFactory({
            authorizationToken: authorizationToken,
            region: 'eastus'
        });
        window.WebChat.renderWebChat({
            directLine: createDirectLine({
                secret: '********************'
            }),

            webSpeechPonyfillFactory: webSpeechPonyfillFactory
        }, document.getElementById('webchat'));
    }
})().catch(err => console.error(err));

Can anyone guide me if there is any way to customize Directline speech for accents.

Upvotes: 0

Views: 143

Answers (1)

Steven Kanberg
Steven Kanberg

Reputation: 6368

Unfortunately, Direct Line Speech does not support accents. The full list of supported languages, including regional variations (for example Spanish (Honduras) vs Spanish (Panama)) can be found here.

Direct Line Speech does support SSML (Speech Synthesis Markup Language) which has a rich set of associated features. Some options that may help you are:

  • Adjusting speaking style to represent mood
  • Adding / removing breaks or pauses
  • Using phonemes to adjust pronunciation
  • Using custom lexicons to adjust pronunciation
  • Adjusting prosody (i.e., pitch, rate, valume, etc.)

One last option is to create a custom neural voice which

lets you create a one-of-a-kind customized synthetic voice for your applications.

In this case, you would provide audio and/or text samples in order to train the voice for use in customizing it.

Hope of help!

Upvotes: 1

Related Questions