Create a custom speech-to-text endpoint

After you have created custom acoustic models or language models, you can deploy them in a custom speech-to-text endpoint.

Create an endpoint

To create a new custom endpoint, select Deployments on the Custom Speech menu at the top of the page. This action takes you to the Deployments page, which contains a table of current custom endpoints. If you have not yet created any endpoints, the table is empty. The current locale is reflected in the table title.

To create a deployment for a different language, select Change Locale. For more information about supported languages, see Supported locales in Custom Speech Service.

To create a new endpoint, select Create New. In the Create Deployment pane, enter information in the Name and Description boxes of your custom deployment.

In the Subscription combo box, select the subscription that you want to use. If it is an S2 subscription, you can select scale units and content logging. For more information about scale units and logging, see Custom Speech Service meters and quotas.

The following table shows how scale units map to available concurrent requests:

Scale unit Number of concurrent requests
0 1
1 5
2 10
3 15
n 5 * n

You can also select whether content logging is switched on or off. That is, you're selecting whether the endpoint traffic is stored for Microsoft internal use.

In addition, Microsoft provides a rough estimate of costs so that you are aware of the impact on the costs of scale units and content logging. This estimate is a rough estimate and might differ from your actual costs.


These settings are not available with F0 (free tier) subscriptions.

In the Acoustic Model list, select the acoustic model that you want, and in the Language Model list, select the language model that you want. The choices for acoustic and language models always include the base Microsoft models. The selection of the base model limits the combinations. You cannot mix conversational base models with search and dictate base models.

The Create Deployment page


Be sure to accept the terms of use and pricing information by selecting the check box.

After you have selected your acoustic and language models, select Create. This action returns you to the Deployments page. The table now includes an entry that corresponds to your new endpoint. The endpoint’s status reflects its current state while it is being created. It can take up to 30 minutes to instantiate a new endpoint with your custom models. When the status of the deployment is Complete, the endpoint is ready for use.

The Deployments page

When the deployment is ready, the deployment name becomes a link. Selecting the link displays the Deployment Information page, which displays the URLs of your custom endpoint to use with either an HTTP request or the Microsoft Cognitive Services Speech Client Library, which uses web sockets.

The Deployment Information page

Next steps

For more tutorials, see: