How to Deploy Custom Neural Voice Models Locally (On-Premises) and Obtain Model Image Files

Mohammed Salama 20

Hello Microsoft Support Team,

I have successfully created and trained two Custom Neural Voice (CNV) models in Azure Cognitive Services. Currently, I can leverage these models via the Azure-hosted endpoints, but my goal is to deploy them locally in an on-premises or containerized environment (e.g., running on Oracle Cloud Infrastructure or another on-prem setup).

My questions are as follows:

Local Deployment of CNV Models: Is there an officially supported method to export, download, or containerize the trained Custom Neural Voice models so they can be run entirely outside of Azure’s hosted environment? I am aware of the general cognitive services containers for standard voices, but I am looking specifically for guidance related to Custom Neural Voice models.

Model Image Files or Artifacts: Do you provide a way to retrieve a model artifact (such as a Docker image, serialized model file, or any equivalent) for the Custom Neural Voices? If not directly, are there any recommended workarounds or preview features that would allow enterprise customers to achieve offline deployments of their CNV models?

Documentation & Best Practices: If direct export is not supported at this time, could you point me to official documentation or resources that provide guidance on how to integrate CNV services in a hybrid manner, possibly through remote calls from on-prem solutions, or any roadmap items for future offline deployment capabilities?

Any information, including confirmation of limitations, available alternatives, or upcoming features that might enable local/container-based deployment of CNV models would be greatly appreciated. Thank you in advance for your assistance!

santoshkc 11,620 Reputation points Microsoft Vendor

2024-12-11T14:07:54.5833333+00:00
Hi @Mohammed Salama,

Thank you for reaching out to Microsoft Q&A forum!

Currently, deploying Azure Custom Neural Voice (CNV) models locally in an on-premises or containerized environment is not supported by Azure Cognitive Services. CNV models are tightly integrated into Azure's hosted environment due to licensing, intellectual property, and computational considerations.

Below are the responses as per your queries:

Microsoft does not support downloading or exporting Custom Neural Voice (CNV) models for local use, as they are managed exclusively in Azure to ensure licensing compliance and optimal performance. For local needs, consider Azure Cognitive Services Containers (for standard voices).

Microsoft does not provide serialized model files, Docker images, or similar artifacts for Custom Neural Voice (CNV) models, primarily due to proprietary technology that relies on Azure's deep-learning infrastructure and licensing restrictions governing voice data usage.

To integrate Custom Neural Voice (CNV) models in hybrid environments, you can use Azure-hosted services by leveraging the REST API or SDKs to make calls to CNV endpoints from on-premises systems. While offline deployment for CNV models is not currently supported, keep an eye on the Azure Updates page for any future announcements regarding this capability.

For more info: What is custom neural voice?

I hope you understand! Thank you.
santoshkc 11,620 Reputation points Microsoft Vendor

2024-12-12T11:20:00.2633333+00:00

Hi @Mohammed Salama,

Following up to see if the given response was helpful. Thank you.
santoshkc 11,620 Reputation points Microsoft Vendor

2024-12-13T11:34:37.09+00:00

Hi @Mohammed Salama,

We haven’t heard from you on the last response and was just checking back to see if the given response was helpful. Thank you.

Share via

How to Deploy Custom Neural Voice Models Locally (On-Premises) and Obtain Model Image Files

Your answer