How to create Custom Speech service in AI Foundry resource, key-based auth disabled

Question

How to create Custom Speech service in AI Foundry resource, key-based auth disabled

Quynh Huynh (NON EA SC ALT) 40 Microsoft Employee

Hi there,

I am facing issues when trying to create a Custom Speech service in my AI Foundry project. I have Cognitive Services User access on the resource and AI Account Owner on the project, but whenever I try to create a Speech service, I get the below error:

Failed to load finetune custom speech task locales AuthenticationTypeDisabled: Key based authentication is disabled for this resource.
AxiosError: Request failed with status code 403

I have tried following steps from this MSLearn article and enabled Entra ID authentication for the Foundry project via connected resources (connected to the Foundry resource): https://learn.microsoft.com/en-us/azure/ai-foundry/foundry-models/how-to/configure-entra-id.

But I still face the same error. Please let me know how to proceed without enabling Key based authentication for the resource.

Anshika Varshney 3,790 Reputation points Microsoft External Staff Moderator

2025-11-07T10:19:17.56+00:00
Hi Quynh Huynh (NON EA SC ALT),

Yes, the recommended flow is create the Speech resource in the Azure Portal first, then connect it to your Azure AI Foundry project, and finally fine‑tune/deploy from Foundry. Below are the exact steps (with links) you can follow end‑to‑end.

Create your Azure AI Speech resource (Azure Portal)

Reference: Speech service entry point and docs overview.

Connect the Speech resource to your Azure AI Foundry project

This is the documented way to “bring an existing Azure AI services resource into a Foundry project.” [EdgeOSIO-2...roceedings | Word]

Create a Custom Speech job (fine‑tune) in Foundry

You can bring models from Speech Studio into Foundry or start directly in Foundry. [Customize...soft Learn | Learn.Microsoft.com

Deploy the fine‑tuned model and use it

See Speech docs for STT quickstarts and API references (Foundry and SDK paths).

[Speech ser...soft Learn | Learn.Microsoft.com]

Please let me know if there are any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Answer accepted by question author

0 additional answers

Your answer

Anshika Varshney 3,790 Reputation points Microsoft External Staff Moderator

2025-11-07T10:19:17.56+00:00

Hi Quynh Huynh (NON EA SC ALT),

Yes, the recommended flow is create the Speech resource in the Azure Portal first, then connect it to your Azure AI Foundry project, and finally fine‑tune/deploy from Foundry. Below are the exact steps (with links) you can follow end‑to‑end.

Create your Azure AI Speech resource (Azure Portal)

Reference: Speech service entry point and docs overview.

Connect the Speech resource to your Azure AI Foundry project

This is the documented way to “bring an existing Azure AI services resource into a Foundry project.” [EdgeOSIO-2...roceedings | Word]

Create a Custom Speech job (fine‑tune) in Foundry

You can bring models from Speech Studio into Foundry or start directly in Foundry. [Customize...soft Learn | Learn.Microsoft.com

Deploy the fine‑tuned model and use it

See Speech docs for STT quickstarts and API references (Foundry and SDK paths).

[Speech ser...soft Learn | Learn.Microsoft.com]

Please let me know if there are any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Answer 1

Hi Quynh Huynh (NON EA SC ALT) ,

Thanks for sharing the details. From your description, it looks like you’re trying to create or connect a Custom Speech resource inside Azure AI Foundry, and you’re seeing the error.

This issue usually occurs when your Speech resource has key-based authentication disabled, but the AI Foundry workflow (or API call) internally attempts to use key-based authentication for validation. While the Azure portal UI can sometimes handle this automatically, the API and fine-tuning setup path within AI Foundry still require access through keys for certain operations, especially when initializing a Custom Speech project or retrieving locales.

You can replicate the issue by creating a Speech resource with only Azure AD (Entra ID) authentication enabled, then trying to start a Custom Speech fine-tuning project via the AI Foundry interface or REST API. The operation will likely fail with a 403 error when the resource doesn’t support key-based authentication.

Follow some Troubleshooting Steps:

1.Check Authentication Settings:

Go to your Speech resource in the Azure Portal → Keys and Endpoint → confirm if key-based authentication is disabled.
Temporarily enable key-based authentication and retry the operation.

2.Verify Region and Access:

Ensure your Speech resource is in a region that supports Custom Speech fine-tuning.
Make sure your user or service principal has the required roles: Cognitive Services User on the Speech resource and AI Project Owner in Foundry.

3.Reattach Resource:

In AI Foundry, disconnect and re-add the Speech resource connection after enabling key-based auth to refresh credentials.

4.Retry Fine-Tuning or Dataset Upload:

After reattaching, try creating the Custom Speech project again or call the API. The operation should now succeed if key-based auth is available.

Reference Documentation:

Create a Custom Speech project – Azure AI Speech
Azure AI Foundry – Connect AI Services I hope this clears things up! Thankyou

Anshika Varshney 3,790 Reputation points Microsoft External Staff Moderator

2025-11-13T08:53:15.5233333+00:00

Hi Quynh Huynh (NON EA SC ALT) ,

I’ve converted my earlier comment into an answer, could you please take a moment to mark it as Accepted? This helps others in the community with the same question find the solution more easily.

Thankyou!

Share via

How to create Custom Speech service in AI Foundry resource, key-based auth disabled

0 additional answers

Your answer