How to let original custom neural voice speak.

Question

I've tried to let my deployed custom neural voice speak according to "Quickstart: Convert text to speech", but it didn't work. The quick start program works normally when using existing voice name like "en-US-JennyNeural" or "ja-JP-NanamiNeural". But when I try my end point like "speechConfig.EndpointId= xxxxxxxxx", it get an error of "Bad Request". Does anyone know how to specify end point parameter and let custom voice speak? PLS let me know.

Answer 1

マイケル山口 Do you see the endpoint URL details in your deploy model pane or tab? It should look like below:

Depending on the region of your resource the base URL should start with https://<region>.voice.speech.microsoft.com/cognitiveservices/v1 and then append the deployment id as a query parameter.

speechConfig.EndpointId= "https://eastus.voice.speech.microsoft.com/cognitiveservices/v1?deploymentId=<your_deploymentid>"

The voice name should be the same as the one used in the studio and should also be updated in your SSML.

Did you get a chance to refer the documentation on using custom voice?