How to connect to an Azure Custom Speech model

Genesys Cloud Integrations
1

Hi,

I have deployed a custom speech model under my Azure Cognitive Services subscription, but i don't see any place to put my custom model endpoint in Genesys Integration for Microsoft Azure STT. It only accepts the subscription key and the issue token endpoint, which does not carry any information about custom model. So how can I use my custom speech model trained in Azure speech studio in Genesys Cloud?

2

Hello,

From a person from engineering - referring to the URI mentioned in:

The Endpoint URI will be different for each custom speech model. Thats the URI to enter
This information should be fetched from from the "Deploy models" in Speech Studio
e.g. wss://eastus.stt.speech.microsoft.com/speech/recognition/conversation/cognitiveservices/v1?cid= where the uuid is the speech model Endpoint ID

Regards,

3

Hi Jerome,

I have the same issue on my side, and custom model is a really important feature for us.

I'm not sure that I understand what you're saying. When calling STT Cognitive Services API, you need a region and subscription key (mandatories).

So, I'm not sure of that

The Endpoint URI will be different for each custom speech model.

And in the documentation it says the following :

Note : Your regional endpoint URI must be in the following format: https://<region>.api.cognitive.microsoft.com/sts/v1.0/issuetoken . Replace with a region supported by Azure Cognitive Services.

and

e. Under Endpoint URI, enter the regional endpoint URI that you receive from Microsoft Azure Cognitive Services.

So that would be this format then for custom model? wss://eastus.stt.speech.microsoft.com/speech/recognition/conversation/cognitiveservices/v1?cid=

Regards,
Philippe

4

Hello,

I don't have an Azure account with access to Cognitive Services, and never tried to create a custom model.

But in the configuration of the Azure Cognitive Services Text to Speech integration (I know your question is about STT), when setting credentials, it shows the following comment for the Endpoint Uri: The endpoint URI for the Microsoft Azure Cognitive Speech service. This will be in the form https://.api.cognitive.microsoft.com/sts/v1.0/issuetoken for non-custom endpoints

So my understanding would be that if you are to use the standard (non custom) Azure endpoint, when configuring the credentials of your integration (Genesys Cloud side), you will set the subscription key and will use https://.api.cognitive.microsoft.com/sts/v1.0/issuetoken for the Endpoint URI (replacing region with your Azure region for this service).

If you use a custom voice model, as Endpoint URI, you would use what appears on Azure side, when deploying your model with Speech Studio. It should give you an endpoint.
What engineering was saying is that this endpoint will be something like https://.stt.speech.microsoft.com/speech/recognition/conversation/cognitiveservices/v1?cid= (maybe not wss as I copied initially)
I assume this URI should appear when adding a deployment endpoint.

I see similar uri and cid is mentioned on this "Speech to text REST API for short audio" page.

Hope this helps.

Regards,

5

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.