Not able to access the custom neural voice trained model through API

Question

Not able to access the custom neural voice trained model through API

Vivek P N 0

Hi,

Im pretty new to the world of Azure, and I have been trying to understand the capabilities of Custom neural voice. After multiple trial and error, I was able to train a model and even deploy it, but I'm fully stuck in trying to generate an audio based on text using the REST API.

I'm on a Free tier now, but within 30 days, so I have some credits left to test out this flow. It would be helpful if I could get some help with the following error that Im seeing after making a call to REST API.

POST https://eastus.tts.speech.microsoft.com/cognitiveservices/v1

Request Body:

VivTrained is the name of the deployed model, and eastus is the Service region.

<speak version='1.0' xml:lang='en-US'>  <voice name='eastus-VivTrained'>    Hello, this is your custom neural voice speaking from Azure.  </voice></speak>

Headers:

Ocp-Apim-Subscription-Key: <Resource key>
Content-Type:application/ssml+xml
X-Microsoft-OutputFormat:audio-16khz-32kbitrate-mono-mp3
User-Agent:curl

Response:

400 Unsupported voice eastus-VivTrained.

1 answer

Your answer

Answer 1

Hello Vivek !

Thank you for posting on Microsoft Learn.

You called the regional TTS endpoint and passed a voice name that endpoint doesn’t know about eastus-VivTrained.

CNV models are accessed either via the custom endpoint URL created when you deploy the model, or by adding a deploymentId query parameter to the voice endpoint. You must also use the exact voice name or alias shown in your deployment (no region prefix).

https://learn.microsoft.com/en-us/azure/ai-services/speech-service/rest-text-to-speech

In Speech studio, do to Custom voice then Deployments then in your deployment, copy the Endpoint URL. It usually looks like:

https://eastus.voice.speech.microsoft.com/cognitiveservices/v1?deploymentId=<YOUR-DEPLOYMENT-ID>

Then call it like this :

curl -X POST \
  "https://eastus.voice.speech.microsoft.com/cognitiveservices/v1?deploymentId=<YOUR-DEPLOYMENT-ID>" \
  -H "Ocp-Apim-Subscription-Key: <YOUR_SPEECH_RESOURCE_KEY>" \
  -H "Content-Type: application/ssml+xml" \
  -H "X-Microsoft-OutputFormat: audio-16khz-32kbitrate-mono-mp3" \
  -H "User-Agent: curl" \
  --data-binary @- <<'SSML'
<speak version="1.0" xml:lang="en-US">
  <!-- Use the EXACT voice name/alias shown in the deployment details and do NOT prefix with region -->
  <voice name="VivTrained">
    Hello, this is your custom neural voice speaking from Azure.
  </voice>
</speak>
SSML

Or, if you prefer to assemble the URL yourself, switch to the voice host and add the deployment id:

https://eastus.voice.speech.microsoft.com/cognitiveservices/v1?deploymentId=<YOUR-DEPLOYMENT-ID>

and do not use eastus.tts.speech.microsoft.com for custom voices. https://learn.microsoft.com/en-us/azure/ai-services/speech-service/rest-text-to-speech

Share via

Not able to access the custom neural voice trained model through API

1 answer

Your answer